THIS POST IS CONTINUED FROM PART 12, BELOW--
OBJECTIVE AI CANNOT HAVE A VISION, IT CANNOT PRIORITIZE, IT CANT GLEAN CONTEXT, IT CANT TELL THE MORAL OF A STORY , IT CANT RECOGNIZE A JOKE, IT CANT DRIVE CHANGE, IT CANNOT INNOVATE, IT CANNOT DO ROOT CAUSE ANALYSIS , IT CANNOT MULTI-TASK, IT CANNOT DETECT SARCASM, IT CANNOT DO DYNAMIC RISK ASSESSMENT , IT IS UNABLE TO REFINE OWN KNOWLEDGE TO WISDOM, IT IS BLIND TO SUBJECTIVITY, IT CANNOT EVALUATE POTENTIAL, IT CANNOT SELF IMPROVE WITH EXPERIENCE, IT DOES NOT UNDERSTAND BASICS OF CAUSE AND EFFECT, IT CANNOT JUDGE SUBJECTIVELY TO VETO/ ABORT, IT CANNOT FOSTER TEAMWORK DUE TO RESTRICTED SCOPE, IT CANNOT MENTOR, IT CANNOT BE CREATIVE, IT CANNOT PATENT AN INVENTION, IT CANNOT SEE THE BIG PICTURE , IT CANNOT FIGURE OUT WHAT IS MORALLY WRONG, IT CAN BE FOOLED EASILY USING DECOYS WHICH CANT FOOL A CHILD, IT IS PRONE TO CATASTROPHIC FORGETTING, IT CANNOT EVEN SET A GOAL …
ON THE CONTRARY IT CAN SPAWN FOUL AND RUTHLESS GLOBAL FRAUD ( CLIMATE CHANGE DUE TO CO2 ) WITH DELIBERATE BLACK BOX ALGORITHMS, JUST FEW AMONG MORE THAN 40 CRITICAL INHERENT DEFICIENCIES.
ON THE CONTRARY IT CAN SPAWN FOUL AND RUTHLESS GLOBAL FRAUD ( CLIMATE CHANGE DUE TO CO2 ) WITH DELIBERATE BLACK BOX ALGORITHMS, JUST FEW AMONG MORE THAN 40 CRITICAL INHERENT DEFICIENCIES.
1
https://ajitvadakayil.blogspot.com/2019/08/what-artificial-intelligence-cannot-do.html
2
https://ajitvadakayil.blogspot.com/2019/10/what-artificial-intelligence-cannot-do.html
3
https://ajitvadakayil.blogspot.com/2019/10/what-artificial-intelligence-cannot-do_29.html
4
https://ajitvadakayil.blogspot.com/2019/11/what-artificial-intelligence-cannot-do.html
5
https://ajitvadakayil.blogspot.com/2019/11/what-artificial-intelligence-cannot-do_4.html
6
https://ajitvadakayil.blogspot.com/2019/11/what-artificial-intelligence-cannot-do_25.html
7
https://ajitvadakayil.blogspot.com/2019/11/what-artificial-intelligence-cannot-do_88.html
8
https://ajitvadakayil.blogspot.com/2019/11/what-artificial-intelligence-cannot-do_15.html
9
https://ajitvadakayil.blogspot.com/2019/12/what-artificial-intelligence-cannot-do_94.html
10
https://ajitvadakayil.blogspot.com/2019/12/what-artificial-intelligence-cannot-do.html
11
https://ajitvadakayil.blogspot.com/2019/12/what-artificial-intelligence-cannot-do_1.html
12
https://ajitvadakayil.blogspot.com/2020/02/what-artificial-intelligence-cannot-do.html
'ADS' (algorithmic decision systems), rely on the analysis of large amounts of personal data to infer correlations or, more generally, to derive information deemed useful to make decisions.
Human intervention in the decision-making may vary, and may even be completely out of the loop in entirely automated systems. In many situations, the impact of the decision on people can be significant, such as access to credit, employment, medical treatment, or judicial sentences, among other things.
Entrusting ADS to make or to influence such decisions raises a variety of ethical, political, legal, or technical issues, where great care must be taken to analyse and address them correctly.
If they are neglected, the expected benefits of these systems may be negated by a variety of different risks for individuals (discrimination, unfair practices, loss of autonomy, etc.), the economy (unfair practices, limited access to markets, etc.), and society as a whole (manipulation, threat to democracy, etc.).
ADS may undermine the fundamental principles of equality, privacy, dignity, autonomy and free will, and may also pose risks related to health, quality of life and physical integrity. That ADS can lead to discrimination has been extensively documented in many areas, such as the judicial system, credit scoring, targeted advertising and employment.
Discrimination may result from different types of biases arising from the training data, technical constraints, or societal or individual biases.
ADS create new 'security vulnerabilities' that can be exploited by people with malicious intent.
Since ADS play a pivotal role in the workings of society, for example in nuclear power stations, smart grids, hospitals and cars, hackers able to compromise these systems have the capacity to cause major damage.
ADS such as those used for predictive policing, may become overwhelming and oppressive. ADS can be misused by states to control people, for example by identifying political opponents.
More generally, interest groups or states may be tempted to use these technologies to control and influence citizen behaviour. These technologies can also be used to distort information to damage the integrity of democratic discourse and the reputation of the government or political leaders.
The two main forms of understandability considered are transparency and explainability:---
Transparency is defined as the availability of the ADS code with its design documentation, parameters and the learning dataset when the ADS relies on machine learning (ML). Transparency does not necessarily mean availability to the public. It also encompasses cases in which the code is disclosed only to specific actors, for example for audit or certification.
Explainability is defined as the availability of explanations about the ADS. In contrast to transparency, explainability requires the delivery of information beyond the ADS itself. Explanations can be of different types (operational, logical or causal); they can be either global (about the whole algorithm) or local (about specific results); and they can take different forms (decision trees, histograms, picture or text highlights, examples, counterexamples, etc.).
The strengths and weaknesses of each explanation mode should be assessed in relation to the recipients of the explanation (e.g. professional or individual), their level of expertise, and their objectives (to challenge a decision, take actions to obtain a decision, verify compliance with legal obligations, etc.).
Accountability is another key desideratum often put forward in the context of ADS. In accordance with previous work in this area, we see accountability as an overarching principle characterised by the obligation to justify one's actions and the risk of sanctions if justifications are inadequate.
Accountability can therefore be seen as a requirement on a process (obligation to provide justification), which applies to both intrinsic and extrinsic requirements for ADS (each case corresponding to specific types of 'justification').
Safety: is an important issue to consider, especially when ADS are embedded in physical systems whose failure may cause fatal damage.
While many ADS failures can be addressed with ad-hoc solutions, there is a strong need to define a unified approach to prevent ADS from causing unintended harm. A minimum requirement should be to perform extensive testing and evaluation before any large-scale deployment. It is also important to provide accountability, including the possibility of independent audits and to ensure a form of human oversight.
Integrity and availability: Increasingly, ADS will be used in critical contexts. It is therefore important to guarantee that they are secure against malicious adversaries. ADS should not jeopardise integrity and availability. Since most ADS rely heavily on machine learning algorithms, it is important to consider their security properties in the context of these algorithms. Adversaries can threaten the integrity or availability of ADS in different ways, i.e., by polluting training datasets with fake data, attacking the machine learning (ML) algorithm itself or exploiting the generated model (the ADS) at run-time.
Confidentiality and privacy: An adversary may seek to compromise the confidentiality of an ADS. For example, they may try to extract information about the training data or retrieve the ADS model itself. These attacks raise privacy concerns as training data is likely to contain personal data. They may also undermine intellectual property since the ADS model and the training data may be proprietary and confidential to the owner. It can involve anonymising the training datasets and the generated models i.e. designing privacy-preserving ADS.
Fairness (absence of undesirable bias): ADS are often based on machine learning algorithms that are trained using collected data. This process includes multiple potential sources of unfairness. Unfair treatment may result from the content of the training data, the way the data is labelled or the feature selection. As shown in this study, there are different definitions of fairness, and others will be proposed in the future.
Many definitions of fairness are actually incompatible. Explainability: Three main approaches can be followed to implement the requirements of explainability:--
The black box approach: this approach analyses the behaviour of the ADS without 'opening the hood', i.e. without any knowledge of its code. Explanations are constructed from observations of the relationships between the inputs and outputs of the system. This is the only possible approach when the operator or provider of the ADS is uncollaborative (does not agree to disclose the code). Example of this category of approach include LIME (local interpretable model-agnostic explanations),
The white box approach: in contrast to the black box approach, this approach assumes that analysis of the ADS code is possible. An example of early work in this direction is the Elvira system for the graphical explanation of Bayesian networks
Two options are possible to achieve explainability by design: (1) relying on an algorithmic technique which, by design, meets the intelligibility requirements while providing sufficient accuracy, or (2) enhancing an accurate algorithm with explanation facilities so that it can generate, in addition to its nominal results (e. g. classification), a faithful and intelligible explanation for these results.
Higher levels of accuracy and precision may reduce intelligibility. In addition, their evaluation is a difficult (and often partly subjective) task.
Legal instruments: Technical solutions are necessary but cannot solve all the issues raised by ADS by themselves. They must be associated with other types of measures and in particular legal requirements for transparency, explainability or accountability.
'ADS' (algorithmic decision systems), systems often rely on the analysis of large amounts of personal data to infer correlations or, more generally, to derive information deemed useful to make decisions.
Human intervention in the decision-making may vary, and may even be completely out of the loop in entirely automated systems. In many situations, the impact of the decision on people can be significant, such as: access to credit, employment, medical treatment, judicial sentences, etc.
Entrusting ADS to make or to influence such decisions raises a variety of different ethical, political, legal, or technical issues, where great care must be taken to analyse and address them correctly. If they are neglected, the expected benefits of these systems may be counterbalanced by the variety of risks for individuals (discrimination, unfair practices, loss of autonomy, etc.), the economy (unfair practices, limited access to markets, etc.) and society as a whole (manipulation, threat to democracy, etc.).
Different requirements such as transparency, explainability, data protection and accountability are often presented as ways to limit these risks but they are generally ill-defined, seldom required by law, and difficult to implement.
Decision-making algorithms are increasingly used in areas such as access to information, e-commerce, recommendation systems, employment, health, justice, policing, banking and insurance.
They also give rise to a variety of risks, such as discrimination, unfairness, manipulation or privacy breaches.
The need to scrutinise the use of algorithms for decision-making and whether algorithmic decisionmaking can be done in a transparent and accountable way.
An algorithm is an unambiguous procedure to solve a problem or a class of problems. It is typically composed of a set of instructions or rulesthat take some input data and return outputs.
An algorithm can be hand-coded, by a programmer, or generated automatically from data, as in machine learning.
Algorithms are harnessing volumes of macro- and micro-data to influence decisions affecting people in a range of tasks
A distinction is sometimes drawn between predictive and prescriptive ADS, but the frontier between the two categories is often fuzzy. The difference between predictive analytics and prescriptive analytics is the outcome of the analysis.
Predictive analytics provides you with the raw material for making informed decisions, while prescriptive analytics provides you with data-backed decision options that you can weigh against one another.
Predictive analytics provides you with the raw material for making informed decisions, while prescriptive analytics provides you with data-backed decision options that you can weigh against one another.
Predictive analytics transforms all the scattered knowledge you have relating to how and why something happened into models, suggesting future actions. By integrating various techniques including data mining, modelling, machine learning (ML) and artificial intelligence (AI), predictive analytics tools transform the data at hand into focused marketing action.
While descriptive analytics helps us learn more about the past, predictive analytics looks into the future, answering the “What will happen if…” questions.
In marketing, that translates to:---
Anticipating consumer behaviours before they happen
Optimising your marketing campaigns around specific factors that are proven to impact sales
Finding the most valuable leads in your CRM and pitching them with offers that will generate the highest conversions
ADS that aim at improving general knowledge or technology: ADS in this class use algorithms to generate new knowledge, generally through the analysis of complex phenomena. Algorithms are crucial in this context since they can be used to analyse very large datasets to extract knowledge.
They can, for example, help improve climate forecasts, detect diseases or discover new viruses.
These ADS are used to make decisions which have a global impact (or an impact on society) rather than on specific individuals.
ADS that aim at improving or developing new digital services: Applications of this category are used to help make predictions, recommendations or decisions in various areas such as information, finance, planning, logistics, etc. These services aim at optimising one or several specific criteria, such as time, energy, cost, relevance of information, etc
ADS integrated within cyber physical systems: Within this context, ADS are used to provide autonomy to physical objects by limiting human supervision. Examples are autonomous cars, robots or weapons. Autonomous cars are being experimented with all over the world.
Algorithms should replace, or at least assist, users in the way they operate vehicles and should make decisions on behalf of 'drivers'. The goals are essentially to make roads safer and optimize connection times. Similarly, autonomous robots are being developed to help or replace humans in performing difficult physical tasks at work or in the home.
Examples include robots used in factory chains, domestic robots that provide services to humans, or robots on the battlefield. A variety of autonomous weapons are under development to assist soldiers in action and to limit collateral damage.
Challenging ADS decisions: Another major issue with opaque ADS is that they make it difficult to challenge a decision based on their results.
IF AI IS USED BY JUDGES IN COURTS, WHY HAVE HUMAN JUDGES.. LET US REPLACE THESE CUNTS BY ROBOTS.. LET US HAVE FUN
Most ADS operate as 'black boxes' and therefore lack transparency, making their efficiency debatable. Since autonomous weapons embed many algorithms, they are prone to cyber-attacks. If they were actually deployed, the risk of malfunctioning, error or misuse should first be carefully addressed.
.
Understanding algorithmic decision-making: . disadvantage is that they can amplify biases and errors and make it more difficult to allocate liabilities.
In contrast with transparency, explainability requires the delivery of information beyond the
ADS itself.:--
– Explanations can be of three different types: operational (informing how the system actually works), logical (informing about the logical relationships between inputs and results) or causal (informing about the causes for the results).
– Explanations can be either global (about the whole algorithm) or local (about specific results).
Accountability is another key desideratum that is often put forward in the context of ADS..
An adversary can threaten the integrity or availability of such ADS in different ways:--
• by attacking the training dataset, for example, by injecting fake data,
• by attacking the ML algorithm, or
• by exploiting the generated model (the ADS) at run-time.
Attackers may want to retrieve some of the data used to train the system. Two main types of scenarios can be considered:--
• 'White box' attacks rely on the assumption that the attacker has access to the model and tries to learn about the training data by 'inverting' it.
• 'Black box' attacks do not assume access to the model: an adversarial client can only submit ueries to the model and make predictions based on the answers.
In contrast to the 'black box' approach, 'white box' explanation systems do rely on the analysis of the ADS code. In addition to the type of explanations that they can generate, 'white box' solutions differ in terms of the ADS they can handle (Bayesian networks, neural networks of limited depth, deep neural networks, etc.), their way to handle continuous data (e.g. through discretisation) and their complexity.
An example of early work in this direction is the Elvira system for the graphical explanation of Bayesian networks.
An example of early work in this direction is the Elvira system for the graphical explanation of Bayesian networks.
An adversary can threaten the integrity and availability of an ADS by polluting its training dataset, attacking its underlying algorithm or exploiting the generated model at run-time.
'Hand-coded' ADS code can be audited, but the task is not always easy since they generally consist of complex modules made of a large number of code lines developed by groups of engineers. ADS that are based on machine learning are even more challenging to understand, and therefore to explain, since their models are generated automatically from training data. Data have many properties and features, and each of them can influence the generated models.
One of the most widely discussed and commented regulations passed during recent years is the European General Data Protection Regulation (GDPR).
In particular, it introduces:--
• new rights for individuals (such as the right to portability, stricter rules for information and consent, enhanced erasure rights, etc.),
• new obligations for data controllers (data protection impact assessments, data protection by design and default, data breach notifications, etc.),
• new action levers such as collective actions and higher sanctions,
• better coordination mechanisms between supervisory authorities and a new body, the European
Data Protection Board (EDPB), which replaces former Article 29 Working Party and which has extensive powers and binding decisions in particular for dispute resolution between national supervisory authorities.
Explaining or detecting biases in ADS should be considered lawful and should not be limited by trade secret or more generally by intellectual property right laws.
The development of a body of experts in ADS, with the ability to cover both technical and ethical aspects, should also be encouraged.
These experts could be integrated into development teams or serve in ADS evaluation bodies.
Because ADS are used to make decisions about people, it is of prime importance that all everyone involved have a minimum of knowledge about the underlying processes, their potential and the limitations of the technologies. As , digital literacy is essential for citizens to be able to exercise their rights in the digital society.
Enhancing the level of understanding of the technologies involved in ADS is necessary, but not sufficient since many issues raised by ADS are subjective and may be approached in different ways depending on individual perceptions and political views.
If an algorithm is designed to preclude individuals from taking responsibility within a decision, then the designer of the algorithm should be held accountable for the ethical implications of the algorithm in use.
In the context of algorithmic decision-making, an accountable decision-maker must provide its decision-subjects with reasons and explanations for the design and operation of its automated decision-making system..
.
Algorithm impact assessments AIAs strive to achieve four initial goals:--
Respect the public’s right to know which systems impact their lives and how they do so by publicly listing and describing algorithmic systems used to make significant decisions affecting identifiable individuals or groups, including their purpose, reach, and potential public impact;
Ensure greater accountability of algorithmic systems by providing a meaningful and ongoing opportunity for external researchers to review, audit, and assess these systems using methods that allow them to identify and detect problems;
Increase public agencies’ internal expertise and capacity to evaluate the systems they procure, so that they can anticipate issues that might raise concerns, such as disparate impacts or due process violations; and
Ensure that the public has a meaningful opportunity to respond to and, if necessary, dispute an agency’s approach to algorithmic accountability. Instilling public trust in government agencies is crucial — if the AIA doesn’t adequately address public concerns, then the agency must be challenged to do better.
Rights become dangerous things if they are unreasonably hard to exercise or ineffective in results, because they give the illusion that something has been done while in fact things are no better.'
Algorithmic Impact Assessments must set forth a reasonable and practical definition of automated decision making In order for AIAs to be effective, agencies must publish their definition as part of a public notice and comment process whereby individuals, communities, researchers, and policymakers could respond, and if necessary challenge, the definition’s scope. This would allow push back when agencies omit essential systems that raise public concerns.
Algorithmic Impact Assessments should provide a comprehensive plan for giving external researchers meaningful access to examine specific systems and gain a fuller account of their workings. Algorithmic Impact Assessments must include an evaluation of how a system might impact the public, and show how they plan to address any issues, should they arise.
Algorithmic Impact Assessment process should provide a path for the public to pursue cases where agencies have failed to comply with the Algorithmic Impact Assessment requirement, or where serious harms are occurring
In many situations, the impact of the decision on people can be significant, such as on access to credit, employment, medical treatment, judicial sentences, among other things. Entrusting ADS to make or to influence such decisions raises a variety of different ethical, political, legal, or technical issues, where great care must be taken to analyse and address them correctly.
If they are neglected, the expected benefits of these systems may be negated by the variety of risks for individuals (discrimination, unfair practices, loss of autonomy, etc.), the economy (unfair practices, limited access to markets, etc.), and society as a whole (manipulation, threat to democracy, etc.).
ADS may undermine the fundamental principles of equality, privacy, dignity, autonomy and free will, and may also pose risksrelated to health, quality of life and physical integrity. That ADS can lead to discrimination has been extensively documented in many areas, such as the judicial system, credit scoring, targeted advertising and employment.
The risk of discrimination related to the use of ADS should be compared with the risk of discrimination without the use of ADS.
.
ADS create new 'security vulnerabilities' that can be exploited by people with malicious intent.
ADS such as those used for predictive policing, may become overwhelming and oppressive like in Israel where even blockchain is used to grab Palestinian land..
ADS can be misused by states to control people, for example by identifying political opponents. More generally, interest groups or states may be tempted to use these technologies to control and influence citizen behaviour. These technologies can also be used to distort information to damage the integrity of democratic discourse and the reputation of the government or political leaders.
.
Transparency is defined as the availability of the ADS code with its design documentation, parameters and the learning dataset when the ADS relies on machine learning (ML).
Transparency does not necessarily mean availability to the public. It also encompasses cases in which the code is disclosed only to specific actors, for example for audit or certification.
.
Explainability is defined as the availability of explanations about the ADS. In contrast to transparency, explainability requires the delivery of information beyond the ADS itself.
Explanations can be of different types (operational, logical or causal); they can be either global (about the whole algorithm) or local (about specific results); and they can take different forms (decision trees, histograms, picture or text highlights, examples, counterexamples, etc.).
The strengths and weaknesses of each explanation mode should be assessed in relation to the recipients of the explanation (e.g. professional or individual), their level of expertise, and their objectives (to challenge a decision, take actions to obtain a decision, verify compliance with legal obligations, etc.)...
Since most ADS rely heavily on machine learning algorithms, it is important to consider their security properties in the context of these algorithms. Adversaries can threaten the integrity or availability of ADS in different ways, i.e., by polluting training datasets with fake data, attacking the machine learning (ML) algorithm itself or exploiting the generated model (the ADS) at run-time. We argue that existing protection mechanisms remain preliminary and require more research.
ADS are often based on machine learning algorithms that are trained using collected data. This process includes multiple potential sources of unfairness. Unfair treatment may result from the content of the training data, the way the data is labeled or the feature selection..
The black box approach: this approach analyses the behaviour of the ADS without 'opening the hood', i.e. without any knowledge of its code. Explanations are constructed from observations of the relationships between the inputs and outputs of the system. This is the only possible approach when the operator or provider of the ADS is uncollaborative (does not agree to disclose the code). Examples of this category of approach include LIME (local interpretable model-agnostic explanations), .
The white box approach: in contrast to the black box approach, this approach assumes that analysis of the ADS code is possible. An example of early work in this direction is the Elvira system for the graphical explanation of Bayesian networks.
The Elvira system is a tool to construct model based decision support systems. The models supported are based on probabilistic uncertainty. It is a tool to construct model based decision support systems.
Bayesian networks are a type of Probabilistic Graphical Model that can be used to build models from data and/or expert opinion. They can be used for a wide range of tasks including prediction, anomaly detection, diagnostics, automated insight, reasoning, time series prediction and decision making under uncertainty.
Other solutions based on neural networks have been proposed more recently.
The constructive approach: in contrast to the first two approaches, which assume that the ADS already exists, the constructive approach is to design ADS taking explainability requirements into account ('explainability by design').
Two options are possible to achieve explainability by design: (1) relying on an algorithmic technique which, by design, meets the intelligibility requirements while providing sufficient accuracy, or (2) enhancing an accurate algorithm with explanation facilities so that it can generate, in addition to its nominal results (e. g. classification), a faithful and intelligible explanation for these results.
Higher levels of accuracy and precision may reduce intelligibility. In addition, their evaluation is a difficult (and often partly subjective) task.
(1) ADS should not be deployed without a prior algorithmic impact assessment (AIA) unless it is clear they have no significant impact on individuals lives; and
(2) the certification of ADS should be mandatory in certain sectors. AIA should not only focus on the risks of using an ADS: they should also assess the risks of not using an ADS.
An algorithm is an unambiguous procedure to solve a problem or a class of problems.
It is typically composed of a set of instructions or rulesthat take some input data and return outputs.
As an example, a sorting algorithm can take a list of numbers and proceed iteratively, first extracting the largest element of the list, then the largest element of the rest of the list, and so on, until the list is empty.
An algorithm can be hand-coded, by a programmer, or generated automatically from data, as in machine learning.
ADS can lead to discrimination has been documented in many areas,such as the justice system, targeted advertisements and employment.
It should be noted that these discriminations do not necessarily arise from deliberate choices: they may result from different types of bias, for example bias in training data (in which case, the algorithm reproduces and systematises already existing discriminations), societal or individual bias (e.g. of designers or programmers of the ADS), or bias arising from technical constraints11 (e.g. limitations of computers or difficulty to formalise the non-formal).
Credit scoring is one of the domains most studied, because the use of ADS in this context can have significant impact on individuals' lives. For The use of certain ADS can also lead to discrimination against underprivileged or minority neighbourhoods.
For example, some geo-navigational applications are designed to avoid 'unsafe neighbourhoods', which could lead to a form of redlining and 'reinforce existing harmful and negative stereotypes about poor communities and communities of colour'.
A major issue with opaque ADS is that they make it difficult to challenge a decision based on their results. This is in contradiction with rights of defence and principles of adversarial proceedings.
The COMPAS (Correctional Offender Management Profiling for Alternative Sanctions) algorithm, which is used by judges to predict whether defendants should be detained or released on bail pending trial, was found to be biased against African-Americans
AI is having an impact on democracy and governance as computerized systems are being deployed to drive objectivity in government functions. algorithms, which are a set of step-by-step instructions that computers follow to perform a task, have become more sophisticated and pervasive tools for automated decision-making
Algorithms are harnessing volumes of macro- and micro-data to influence decisions affecting people in a range of tasks Bias in algorithms can emanate from unrepresentative or incomplete training data or the reliance on flawed information that reflects historical inequalities.
If left unchecked, biased algorithms can lead to decisions which can have a collective, disparate impact on certain groups of people even without the programmer’s intention to discriminate. The exploration of the intended and unintended consequences of algorithms is necessary
Algorithms with too much data, or an over-representation, can skew the decision toward a particular result.
BUT HEY, WHITE JEWS HAVE A NEAR MONOPOLY ON SIGNATURE BASED SADISTIC SERIAL KILLING ( LIKE OF JEW TED KACZYNSKI ) ..
BUT THIS WILL NEVER FIND ITS WAY INTO ADS.. SUCH IS THE STRANGLEHOLD OF JEWS IN AI JUDICIAL SYSTEMS
The point is that most ADS used in this context are risk-assessment tools: based on a number of factors about the defendants' criminal history, sociological data or demographic features, they provide an estimation of their risk of recidivism.
As a result, they privilege one objective (incapacitation, defined as prevention from reoffending) to the detriment of other traditional justifications of punishment in law, such as retribution (taking into account the severity of the crime), rehabilitation (social reintegration) and deterrence.
ADS can contribute to making administration decisions more efficient, transparent and accountable, provided however that they are themselves transparent and accountable.
Most ADS operate as 'black boxes' and therefore lack transparency, making their efficiency debatable.
Existing machine learning technologies enable a high degree of automation in labour-intensive activities such as satellite imagery analysis. A more ambitious and controversial use of ADS in this context is to build autonomous weapon systems. A number of countries are increasing their studies and development of such systems as they perform increasingly elaborate functions, including identifying and killing targets ( using drones ) with little or no human oversight or control.
The results of ADS are often difficult to explain. This can reduce consumer trust and creates four main risks:--
• There may be 'hidden' biases derived from the data provided to train the system. This can be difficult to detect and correct. In some cases, these biases can be characterised as discriminations and be sanctioned in court.
• It can be difficult, if not impossible, to prove that the system will always provide correct outputs, especially for scenarios that were not represented in the training data. This lack of verifiability can be a concern in mission-critical applications.
• In case of failure, it might be very difficult, given the models' complexity, to diagnose and correct the errors and to establish responsibilities. .
• Finally, as previously mentioned, malicious adversaries can potentially attack the systems by poisoning the training data or identifying adversarial examples. These attacks can be difficult to detect and prevent.
ADS can be audited systematically. Their they can amplify biases and errors and make it more difficult to allocate liabilities.
Several explanation modes can be distinguished:--
-- Explanations can be of three different types: operational (informing how the system actually works), logical (informing about the logical relationships between inputs and results) or causal (informing about the causes for the results).
– Explanations can be either global (about the whole algorithm) or local (about specific results).
– Explanations can take different forms (decision trees, histograms, picture or text highlights, examples, counterexamples, etc.).
The strengths and weaknesses of each explanation mode should be assessed in relation to the recipients of the explanations (e.g. professional or individual), their level of expertise and their objectives (understanding the results to make a decision, challenging a decision, verifying compliance with legal obligations, etc.).
Acountability can be seen as a requirement on a process (obligation to provide justification),
Many papers use terms such as transparency, explainability, interpretability, accountability or fairness with different meanings or without defining them properly (and often without introducing clear distinctions between them).
In data mining and machine learning, interpretability is defined as the ability to explain or to provide the meaning in understandable terms to a human. These definitions assume implicitly that the concepts expressed in the understandable terms composing an explanation are self-contained and do not need further explanations.
Interpretability typically means that the model can be explained, a quality which is imperative in almost all real applications where a human is responsible for consequences of the model.'
ADS need to be trained to be able to solve complex tasks.
For example, the cleaning robot should learn to handle candy wrappers differently from a dropped diamond ring
An adversary can threaten the integrity or availability of such ADS in different ways:--
• by attacking the training dataset, for example, by injecting fake data,
• by attacking the ML algorithm, or
• by exploiting the generated model (the ADS) at run-time.
The attacks on the ML algorithm, sometimes called 'logic attacks', require the adversary to have physical access to the systems where the algorithm is running. These attacks are not specific to ADS and can be mitigated by various security measures, such as access control or hardware security.
.
The goal of an attack on the training phase is to influence the generated model by compromising its integrity or availability. Integrity attacks alter the generated model towards a specific goal, for example to maliciously obtain a loan or to go through an intrusion detection system (IDS).
ML classifier, the goal of an integrity attack could be to assign an incorrect class to a legitimate input.
In contrast, availability attacks tend to affect the quality, performance or access to the system. The final goal may be to create sufficient errors to make the ADS unusable.
Although their goals are different, these attacks are similar in nature and are typically performed by altering or poisoning the training dataset by injecting adversarial data (injection attacks) or by removing or modifying existing records (modification attacks). The modification can be performed, in a supervised setting, by modifying the data labels99 or the data itself.
Note that these attacks require that the adversaries have access to the pre-processed training dataset. If this is not possible, the adversary can poison or inject the training data before pre-processing.
Attacks on the execution phase do not intend to modify the ADS generated model, but instead seek to exploit some of its weaknesses. The idea is to compute some inputs, called adversarial examples, which will trigger the desired, incorrect, outputs.
When the ADS is a classifier, the adversary seeks to have the perturbed inputs assigned to incorrect classes.
Most of the previous attacks assume 'white box' scenarios, in which attackers have access to the internal workings of the model. However, the 'black box' scenario is probably a more realistic threat model.
For example, an attacker who wants to attack an image recognition system or a spam filter rarely has access to the internals of the model. Instead, they often have access to the system as an oracle, i.e. it can query the ADS with their own inputs and can observe the generated outputs.
Attacks on 'black box' systems, also called 'black box' attacks, are more challenging but not impossible. A key property in this respect is adversarial example transferability, i.e. the property that can be exploited whereby adversarial examples crafted for a given classifier are likely to be misclassified by other models trained for the same task.
An adversary may want to compromise the confidentiality of an ADS for example by trying to extract information about the training data or by retrieving the ADS model itself. These attacks raise privacy oncerns, since training data often contain personal data.
They may also undermine intellectual property, as the ADS model and the training data can be proprietary and confidential to their owner.
Attackers may want to retrieve some of the data used to train the system.
Two main types of scenarios can be considered:--
• 'White box' attacks rely on the assumption that the attacker has access to the model and tries to learn about the training data by 'inverting' it.
• 'Black box' attacks do not assume access to the model: an adversarial client can only submit queries to the model and make predictions based on the answers.
ADS are often based on machine learning algorithms trained on collected data. There are multiple potential sources of unfairness in this process.
Unfair treatment can result, for example, from the content of the training data, the way the data is labelled or the feature selection.
Biased training data. If the training data contains biases or historical discriminations, the ADS will inherit them and incorporate them into its future decisions.
ADS, and more generally machine learning algorithms, are systems trained to recognise and leverage statistical patterns in data. However, they are not perfect and perform classification or prediction errors.
The accuracy rate of an ADS is often related to the size of the training dataset--large training dataset leads to less errors, and less data leads to worse predictions. ADS are often complex systems that are difficult to understand. 'Hand-coded' ADS code can be audited, but the task is not always easy since they generally consist of complex modules made of a large number of code lines developed by groups of engineers.
ADS that are based on machine learning are even more challenging to understand, and therefore to explain, since their models are generated automatically from training data. Data have many properties and features, and each of them can influence the generated models.
Bias is ‘an inclination of prejudice towards or against a person, object, or position’
Some algorithms collect their own data based on human-selected criteria, which can also reflect the bias of human designers.
Known as 'ADS' (algorithmic decision systems), these systems often rely on the analysis of large amounts of personal data to infer correlations or, more generally, to derive information deemed useful to make decisions.
Algorithmic bias describes systematic and repeatable errors in a computer system that create unfair outcomes, such as privileging one arbitrary group of users over others. Bias can emerge due to many factors, including but not limited to the design of the algorithm or the unintended or unanticipated use or decisions relating to the way data is coded, collected, selected or used to train the algorithm.
Algorithmic bias is found across platforms, including but not limited to search engine results and social media platforms, and can have impacts ranging from inadvertent privacy violations to reinforcing social biases of race, gender, sexuality, and ethnicity. The study of algorithmic bias is most concerned with algorithms that reflect "systematic and unfair" discrimination.
This bias has only recently been addressed in legal frameworks, such as the 2018 European Union's General Data Protection Regulation. Bias can enter into algorithmic systems as a result of pre-existing cultural, social, or institutional expectations; because of technical limitations of their design; or by being used in unanticipated contexts or by audiences who are not considered in the software's initial design.
Commercial algorithms are proprietary, and may be treated as trade secrets.Treating algorithms as trade secrets protects companies, such as search engines, where a transparent algorithm might reveal tactics to manipulate search rankings. This makes it difficult for researchers to conduct interviews or analysis to discover how algorithms function. such secrecy can also obscure possible unethical methods used in producing or processing algorithmic output
The General Data Protection Regulation (GDPR), the European Union's revised data protection regime that was implemented in 2018, addresses "Automated individual decision-making, including profiling" in Article 22. These rules prohibit "solely" automated decisions which have a "significant" or "legal" effect on an individual, unless they are explicitly authorised by consent, contract, or member state law. Where they are permitted, there must be safeguards in place, such as a right to a human-in-the-loop, and a non-binding right to an explanation of decisions reached. While these regulations are commonly considered to be new, nearly identical provisions have existed across Europe since 1995, in Article 15 of the Data Protection Directive.
The United States has no general legislation controlling algorithmic bias, approaching the problem through various state and federal laws that might vary by industry, sector, and by how an algorithm is used. Many policies are self-enforced or controlled by the Federal Trade Commission.In 2016, the Obama administration released the National Artificial Intelligence Research and Development Strategic Plan,which was intended to guide policymakers toward a critical assessment of algorithms. It recommended researchers to "design these systems so that their actions and decision-making are transparent and easily interpretable by humans, and thus can be examined for any bias they may contain, rather than just learning and repeating these biases". Intended only as guidance, the report did not create any legal precedent.
The “Algorithmic Accountability Act of 2019” was introduced in the U.S. House of Representatives on April 10, 2019 and referred to the House Committee on Energy and Commerce. The bill requires an assessment of the risks posed by automated decision systems to the privacy or security of personal information of consumers and the risks that the systems may result in or contribute to inaccurate, unfair, biased or discriminatory decisions impacting consumers.
Governance and accountability issues relate to who creates the ethics standards for AI, who governs the AI system and data, who maintains the internal controls over the data and who is accountable when unethical practices are identified. The internal auditors have an important role to play in this regard. They should assess risk, determine compliance with regulations and report their findings directly to the audit committee of the board of directors.
On July 31, 2018, a draft of the Personal Data Bill was presented in India The draft proposes standards for the storage, processing and transmission of data. While it does not use the term algorithm, it makes for provisions for "...harm resulting from any processing or any kind of processing undertaken by the fiduciary". It defines "any denial or withdrawal of a service, benefit or good resulting from an evaluative decision about the data principal" or "any discriminatory treatment" as a source of harm that could arise from improper use of data. It also makes special provisions for people of "Intersex status
Bias can be introduced in many ways, including the following.--
• It can be present in the (training/validation/test) input dataset. For instance, a common form of bias is human bias (when data are labelled according to a person’s own view and therefore reflect the bias of that person).
• It can be introduced via the online learning process, when new, biased data are fed to the model in real time.
• Bias may also occur when ML models make non-linear connections between disparate data sources, despite those sources being validated individually for certain characteristics/variables.
• It can be introduced into the model during the development phase through inadvertent coding of biased rules, for example (algorithmic bias).
Most commonly, data contain bias when they are not representative of the population in question.
This can lead to discrimination, for example when a class of people less represented in the training dataset receives less or more favourable outcomes simply because the system has learnt from only a few examples and is not able to generalise correctly. However, discrimination can exist without bias or direct discrimination; it can result from sensitive attributes serving as input variables, regardless of bias.
Techniques exist to prevent or detect bias (active or passive de-biasing). For example, controls can be implemented during the data preparation and feature engineering phases to prevent or detect bias and discrimination.
Furthermore, statistical analysis (e.g. data skewness analysis) can be applied to the training dataset to verify that the different classes of the target population are equally represented (under-represented classes can be incremented by oversampling or overrepresented classes can be reduced in size). In addition, techniques (and libraries) exist to test models against discriminatory behaviour (e.g. using crafted test datasets that could lead to discrimination).
5 most common types of bias:---
Confirmation bias. Occurs when the person performing the data analysis wants to prove a predetermined assumption. ...
Selection bias. This occurs when data is selected subjectively. ...
Outliers. An outlier is an extreme data value. ...
Overfitting and underfitting. ...
Confounding variables
A confounding variable is an “extra” variable that you didn't account for. They can ruin an experiment and give you useless results.
In a distribution of variables, outliers lie far from the majority of the other data points as the corresponding values are extreme or abnormal. The outliers contained in sample data introduce bias into statistical estimates such as mean values, leading to under- or over-estimated resulting values . An outlier is a value that escapes normality and can (and probably will) cause anomalies in the results obtained through algorithms and analytical systems
Usually, the presence of an outlier indicates some sort of problem. An outlier may be due to variability in the measurement or it may indicate experimental error; the latter are sometimes excluded from the data set. ... Outliers can occur by chance in any distribution, but they often indicate either measurement error or that the population has a heavy-tailed distribution..
Confirmation bias is the tendency to search for, interpret, favor, and recall information in a way that confirms or strengthens one's prior personal beliefs or hypotheses
Bias can happen in ways that do not involve humans in the loop. In fact, algorithms used for inferencing or training may be working perfectly well and not recognize bias because those aberrations are so small that they escape detection.
But bias also can be cumulative, and in some cases exponential. As such, it can cause issues much further down the line, making it difficult to trace back to where the problem originated.
Biasing can start wherever that data is generated.
Every measurement has accuracy levels and tolerances, but they shift over time. Sensors have variability, and that variability changes over time. So you have to figure out where it is in the calibration cycle and keep trying to correct for variability.
That’s part of the picture. But every piece of data has some variability in it, too. So if you add in all of that data randomly, you could multiply the variability
Recommendations for building a robust and responsive AI and data ethics capability: --
“Appoint chief data/AI officers with ethics as part of their responsibilities.”
“Assemble organizationally high-level ethics advisory groups.”
“Incorporate privacy and ethics-oriented risk and liability assessments into decision-making or governance structures.”
“Provide training and guidelines on responsible data practices for employees.”
“Develop tools, organizational practices/structures, or incentives to encourage employees to identify potentially problematic data practices or uses.”
“Use a data certification system or AI auditing system that assesses data sourcing and AI use according to clear standards.”
“Include members responsible for representing legal, ethical, and social perspectives on technology research and project teams.”
“Create ethics committees that can provide guidance not only on data policy, but also on concrete decisions regarding collection, sharing, and use of data and AI.”
An AI ethics committee should seek to address the following concerns:--
“Whether the project under review advances organizational aims and foundational values to an extent that it justifies any organizational and social risks or costs.”
“Whether the project is likely to violate any hard constraints, such as legal requirements or fundamental organizational commitments/principles.”
“Whether an impartial citizen would judge that the organization has done due diligence in considering the ethical implications of the project.”
“Whether it is possible to secure the sought benefits in a way that better aligns with organizational values and commitments and without any significant additional undue burden or costs.”
“Whether reputational risks could be significant enough to damage the brand value in the concerned market or in other places where the organization operates.”
Tesla and SpaceX founder Elon Musk issued a warning: “Mark my words, AI is far more dangerous than nukes.” “Unless we learn how to prepare for, and avoid, the potential risks,AI could be the worst event in the history of our civilization.”
AI has the potential to hurt people in mass numbers, which puts unique responsibilities on the field. This incredible power to do harm at scale means those of us in the AI industry have a responsibility to put societal interest above profit.
Malicious use of AI, could threaten digital security (e.g. through criminals training machines to hack or socially engineer victims at human or superhuman levels of performance), physical security (e.g. non-state actors weaponizing consumer drones), and political security (e.g. through privacy-eliminating surveillance, profiling, and repression, or through automated and targeted disinformation campaigns)
Digitization is a building block toward artificial intelligence because it can facilitate the availability of the “Big Data” on which machine learning is based. Next on the spectrum would be for governments to rely on what we call here algorithmic tools—that is, traditional, human-created statistical models, indices, or scoring systems that are then used as decision tools.
These traditional algorithmic or statistical tools rely on humans to select the specific variables to be included in a decision aid and the precise mathematical relationships between those variables. Only the final step on the spectrum—machine learning— constitutes what we will consider artificial intelligence, because learning algorithms essentially work “on their own” to process data and discover optimal mathematical relationships between them.
This autonomous self-discovery is what gives machine-learning algorithms not only their name but also their frequent superior performance in terms of accuracy over traditional algorithmic tools. Of course, even with machine learning, humans must specify the objective that the learning algorithm is supposed to forecast or optimize, and humans must undertake a number of steps to “train” the algorithm and refine its operation.
Yet these learning algorithms are different than traditional statistical tools because the precise ways that data are combined and analyzed are neither determined in advance by a human analyst nor easily explainable after the fact.
For this reason, machine learning algorithms are often described as “black-box” algorithms because they do not afford a ready way of characterizing how they work—other than that they can be quite accurate in achieving the objectives they have been designed to achieve.
Biased algorithms can lead to decisions which can have a collective, disparate impact on certain groups of people .
AI is still an emerging technology. Data is analyzed through deep uses of algorithmic programming.
There are lots of examples where the data and humans building the algorithms had an inherent bias that was built into the AI algorithm. In Israel Palestinians are always at the receiving end in a most slimy and evil manner..
Israel was the first to screw a part of the population ( Palestinians ) with blockchain by grabbing their ancestral lands.
Israel is the first to screw a part of its citizens ( Palestinians ) using black box algorithms for justice
Most AI algorithms are essentially black box pattern detectors. They can detect specific patterns quickly, but they don’t give you the causality of why those patterns happen. We can’t see why an algorithm has taken a decision; it’s not explained in a way that can be understood by a human.
And we are making decisions based on these ‘black boxes’ without knowing why. We are losing control over the way decisions are being made – and this is a major issue.
An AI model is considered to be traceable if (a) its decisions, and b) the datasets and processes that yield the AI model’s decision (including those of data gathering, data labelling and the lgorithms used), are documented in an easily understandable way.
Augmented intelligence unites the strengths of people and machines when prospecting value from data. Namely, you can augment human instinct with smart algorithms that provide fast, data-driven predictive insights. These insights can help people redesign functions, detect patterns, find strategic opportunities, and turn data into action.
Decision-making and actions will improve, provided you have feedback loops built in for continuous improvement. Feedback loops are important for improving upon algorithms and in making sure that when things do not happen as expected, there are mechanisms in place to understand why.
Auditability refers to the readiness of an AI system to undergo an assessment of its algorithms, data and design processes
To facilitate auditability, organisations
can consider keeping a comprehensive record of data provenance, procurement, preprocessing, lineage, storage and security. The record could also include qualitative input about data representations, data sufficiency, source integrity, data timelines, data relevance, and unforeseen data issues encountered across the workflow.
1. Accountability: Ensure that AI actors are responsible and accountable for the proper functioning of AI systems and for the respect of AI ethics and principles, based on their roles, the context, and consistency with the state of art.
2. Accuracy: Identify, log, and articulate sources of error and uncertainty throughout the algorithm and its data sources so that expected and worst-case implications can be understood and can inform mitigation procedures.
3. Auditability: Enable interested third parties to probe, understand, and review the behaviour of the algorithm through disclosure of information that enables monitoring, checking or criticism.
4. Explainability: Ensure that automated and algorithmic decisions and any associated data driving those decisions can be explained to end-users and other stakeholders in non-technical terms.
5. Fairness:
a. Ensure that algorithmic decisions do not create discriminatory or unjust impacts across different demographic lines (e.g. race, sex, etc.).
b. To develop and include monitoring and accounting mechanisms to avoid unintentional discrimination when implementing decision-making systems.
c. To consult a diversity of voices and demographics when developing systems, applications and algorithms.
Build trust by ensuring that designers and operators are responsible and accountable for their systems, applications and algorithms, and to ensure that such systems, applications and algorithms operate in a transparent and fair manner.
To make available externally visible and impartial avenues of redress for adverse individual or societal effects of an algorithmic decision system, and to designate a role to a person or office who is responsible for the timely remedy of such issues.
Algorithm audits are conducted if it is necessary to discover the actual operations of algorithms comprised in models. This would have to be carried out at the request of a regulator (as part of a forensic investigation) having jurisdiction over the organisation or by an AI technology provider to assist its customer organisation which has to respond to a regulator’s request.
Conducting an algorithm audit requires technical expertise which may require engaging external experts. The audit report may be beyond the understanding of most individuals and organisations.
The expense and time required to conduct an algorithm audit should be weighed against the expected benefits obtained from the audit report. Ultimately, algorithm audits should normally be used when it is reasonably clear that such an audit will yield clear benefits for an investigation.
Explainability is just one element of transparency. Transparency consists in making data, features, algorithms and training methods available for external inspection and constitutes a basis for building trustworthy models.
Online Dispute Resolution (ODR) has arisen in recent years as a tool for resolving disagreements among parties using technology, growing in part out of prior developments in the field of Alternative Dispute Resolution (ADR). ADR is a term that refers to a range of methods such as mediation and arbitration that aim to settle disputes without the use of litigation and the court system.
eBay and PayPal have developed ODR systems to handle the millions of disputes that regularly arise on their platforms from and among users.
Online dispute resolution (ODR) is a branch of dispute resolution which uses technology to facilitate the resolution of disputes between parties. It primarily involves negotiation, mediation or arbitration, or a combination of all three.
In this respect it is often seen as being the online equivalent of alternative dispute resolution (ADR). However, ODR can also augment these traditional means of resolving disputes by applying innovative techniques and online technologies to the process.
ODR is a wide field, which may be applied to a range of disputes; from interpersonal disputes including consumer to consumer disputes (C2C) or marital separation; to court disputes and interstate conflict
Today, the term ODR is even more expansive. In its current form, ODR covers the use of any technology to assist parties in the dispute resolution process. Consider three different hypothetical mediation proceedings:
A mediation conducted through a video conferencing service rather than in person.
An in-person mediation where the parties utilize technology tools to assist in analyzing their positions.
A mediation run entirely by an artificial intelligence (AI) service.
All three mediations are a form of ODR. ODR can be as straightforward as using a webcam, or as complex as a machine learning algorithm that guides disputants to an optimal settlement. Even the examples above merely scratch the surface of what is possible through ODR. As legal technology improves and expands, ODR is becoming increasingly popular and useful.
key advantages and challenges that face all forms of ODR. First, the advantages:--
Reduced Cost—Cost-saving is already a core advantage of alternative dispute resolution. Instead of engaging in costly and time-intensive litigation, ADR allows parties to minimize costs and save time in resolving their disputes. ODR enhances these benefits. Using telecommunication technology, for instance, eliminates travel expenses, allows for quicker communication, and gives the parties more flexibility in scheduling. Technology tools and AI can help parties understand their position and range of options, leading to more efficient resolution. As a set of tools, ODR can reduce costs significantly as compared to in-person ADR.
Increased Access to Justice—ODR allows for increased access to the legal system in multiple ways. As discussed, ODR originated as a way to address the high number of disputes that arose out of e-commerce transactions. Instead of relying on traditional forms of dispute resolution, companies have incorporated ODR systems into their websites in order to give customers a direct and efficient way to resolve their disputes. For high-volume and low-value disputes, ODR may be the only practical means for a consumer to resolve a dispute.
ODR also allows increased access to the legal system for traditional disputes. A direct byproduct of reduced costs is that more parties can afford to utilize ODR. Courts have begun to implement ODR tools as well. The convenience of ODR makes it significantly easier for users to engage with the court system.
ODR also allows increased access to the legal system for traditional disputes. A direct byproduct of reduced costs is that more parties can afford to utilize ODR. Courts have begun to implement ODR tools as well. The convenience of ODR makes it significantly easier for users to engage with the court system.
Accuracy—The increased convenience of ODR can also lead to better accuracy. Most simply, ODR helps parties and decision-makers reach more accurate outcomes by providing better access to information. ODR can also help avoid implicit biases on factors such as race and socioeconomic status. One goal of AI is to avoid the inevitable biases that are present in human decision-making. Although these tools are still in their infancy, there is hope that ODR can provide more equitable remedies.
ODR faces several challenges as well: --
Fairness—Many of the advantages of ODR are not foolproof. For example, although ODR systems provided by e-commerce companies may increase consumer access to remedies, these systems almost certainly lead to better results to the company paying for them. Although ODR can be more cost-effective, these systems are not costless. Since there is currently no regulation on these systems, outcomes may become stacked in favor of the implementing party.
It is important to pay attention to who is covering the costs of the system, as they are the most likely to benefit from them. Similarly, ODR is not a silver bullet to provide optimally accurate resolutions. For instance, AI can actually exacerbate issues of implicit bias. Developers of these systems need to continually monitor the results to ensure that ODR tools are not creating more inequities than they are solving.
It is important to pay attention to who is covering the costs of the system, as they are the most likely to benefit from them. Similarly, ODR is not a silver bullet to provide optimally accurate resolutions. For instance, AI can actually exacerbate issues of implicit bias. Developers of these systems need to continually monitor the results to ensure that ODR tools are not creating more inequities than they are solving.
Privacy and Security—The introduction of technology inevitably introduces privacy and security risks. No tool is foolproof and therefore information shared through ODR solutions may be at higher risk of exposure than with traditional in-person ADR. Trust is essential in order for parties to resolve a dispute, and that trust extends to the tools and processes they are utilizing. Developers of ODR solutions need to pay particular attention to these concerns.
Impersonality—Disputes are an inherently emotional and trying experience. Mediators and arbitrators do more than act as robots crunching information and outputting settlement ranges. The neutral third-party has to navigate a complex emotional setting and provide an environment to help both parties feel comfortable with the proposed solutions. The benefits of human interaction can be lost as ODR increasingly relies on technology. This risk is particular salient with AI solutions.
Amazon has developed algorithms that can resolve a consumer complaint about a defective product without requiring any human intervention.
Realizing that they could not afford to hire enough human mediators to resolve all of these disputes or arrange for parties to video conference with each other, these companies leveraged the extensive amounts of data they had collected on consumer behavior and usage.
Their ODR systems aim to prevent or amicably resolve as many disputes as possible and to decide the remainder quickly. To do so, they generally first diagnose the problem, working directly with the complainant; they then move to direct negotiations (aided by technology) and ultimately allow the company to decide the case if the parties are not able to amicably resolve matters on their own.
As the success of these systems inspired other firms to develop similar and increasingly sophisticated programs, algorithms have become a more prominent dispute resolution solution, allowing companies to automate away many (if not all) of the steps of decision-making process.
Some courts have also begun experimenting with ODR as a mechanism to attempt to resolve lawsuits without requiring the use of judicial decision-making adopted some form of “court ODR” in cases involving small claim civil matters, traffic violations, outstanding warrant cases, and low-conflict family court cases..
What counts as an ODR system can vary from a simple website that facilitates entering pleas for traffic tickets online to an online portal for engaging in asynchronous negotiations. These are not mandatory systems in any jurisdiction of which we are aware, but instead they are offered as an option to avoid appearing in court. In jurisdictions with these systems, parties are notified of the ODR option via mailings or websites.
Parties can access the ODR system at any time, and with the more interactive systems they can communicate and negotiate with each other, obtain legal information and suggested resolutions from the system, and easily manage electronic documents—all without having to see the inside of a courtroom.
These systems can usually reach resolution in a dispute faster and at lower cost to the parties and are far more accessible than traditional court-centered adjudication. ODR provides an emerging avenue for litigants and courts to engage in dispute resolution outside of the presence of a courtroom and absent a human judge.
Court ODR systems, as well as the private-sector iterations that inspired them, have increasingly adopted automated processes and rely on algorithmic tools to aid in reaching what some observers characterize as fair and low-cost solutions to the parties’ disputes.
Court systems could take these algorithms to the next “level” of autonomy by integrating artificial intelligence into ODR processes, allowing for increasingly automated forms of decision-making for petty ego related cases –like when a rich celebrity ( like a crying bollywood superstar ) wants to harass a desh bhakt man using his poodles in police, whom he has befriended during the annual tamasha named Umang.
COMPAS, an acronym for Correctional Offender Management Profiling for Alternative Sanctions, is a case management and decision support tool used by U.S. courts to assess the likelihood of a defendant becoming a recidivist. Israel uses a similar RA tool to screw Palestinians and grab their ancestral lands.
The COMPAS software uses an algorithm to assess potential recidivism risk. Northpointe created risk scales for general and violent recidivism, and for pretrial misconduct. According to the COMPAS Practitioner's Guide, the scales were designed using behavioral and psychological constructs "of very high relevance to recidivism and criminal careers."
Pretrial Release Risk scale: Pretrial risk is a measure of the potential for an individual to fail to appear and/or to commit new felonies while on release. According to the research that informed the creation of the scale, "current charges, pending charges, prior arrest history, previous pretrial failure, residential stability, employment status, community ties, and substance abuse" are the most significant indicators affecting pretrial risk scores.
General Recidivism scale: The General Recidivism scale is designed to predict new offenses upon release, and after the COMPAS assessment is given. The scale uses an individual's criminal history and associates, drug involvement, and indications of juvenile delinquency.
Violent Recidivism scale: The Violent Recidivism score is meant to predict violent offenses following release. The scale uses data or indicators that include a person's "history of violence, history of non-compliance, vocational/educational problems, the person’s age-at-intake and the person’s age-at-first- arrest." An individual's risk score for violent recidivism is calculated as follows:
Violent Recidivism Risk Score = (age∗−w)+(age-at-first-arrest∗−w)+(history of violence∗w) + (vocation education ∗w) + (history of noncompliance ∗w), where w is weight, the size of which is "determined by the strength of the item’s relationship to person offense recidivism that we observed in our study data."..
This objective bullshit is to keep white Jews safe as Rothschild’s history states that Jews are always at the receiving end— when in reality they are the worst criminals.
As of today, of course, we know of no machine-learning tool that has been adopted in any court in the United States to make an ultimate, fully automated determination on a legal or factual question.
However, several trends in recent years have emerged that could signal movement towards the eventual use of such automated adjudication via artificial intelligence. To date, the principal building blocks of artificial intelligence in the courts comprise the digitization of court filings and processes, the introduction of algorithmic tools for certain criminal court decisions, and the emergence of online dispute resolution as an alternative to traditional court proceedings for small claims.
Some courts have created “dedicated computer kiosks” specifically designed to help litigants who lack legal representation. In California, for example, an “‘Online Self-Help Center’ offers PDFs that can be filled in online and used for evictions, divorces, orders of protection, collection matters, small claims, and other issues.”
The federal judiciary has instituted a “comprehensive case management system” known as the Case Management/Electronic Case Files (CM/ECF) system that allows for convenient filing and organization of court documents, party pleadings, and other relevant materials.
CM/ECF (Case Management/Electronic Case Files) is the case management and electronic court filing system for most of the United States Federal Courts. PACER, an acronym for Public Access to Court Electronic Records, is an interface to the same system for public use.
Public Access to Court Electronic Records (PACER(link is external)) is an electronic public access service that allows users to obtain case and docket information from federal appellate, district, and bankruptcy courts. If you want online access to documents filed in Central District cases, you must have a PACER account.
CM/ECF provides more functionality than PACER, including the ability to electronically file cases and documents, to control electronic service and notice of documents, and to update a user’s contact information for the electronic service of documents
At law firms, the increasing use of algorithmic tools, including those involving machine-learning algorithms, can be found to support the review of documents during the discovery process. This “e-discovery” practice has been shown to have a “strong impact” on reducing the need for human labor—plus it has spawned services that seek to analyze trends and make legal forecasts.
Algorithmic tools have taken root in some court systems as an aid to judicial decision-making in criminal cases on questions of bail, sentencing, and parole— but so far virtually none of these appear to rely on machine-learning algorithms.
No one knows exactly how COMPAS works; its manufacturer refuses to disclose the proprietary algorithm. We only know the final risk assessment score it spits out . . .
Something about this story is fundamentally wrong: Why are we allowing a computer program, into which no one in the criminal justice system has any insight, to play a role in sending a man to prison?
The courts have not yet started to grapple with the legal implications of these algorithmic tools.
ML models can quickly become “black boxes”, opaque systems for which the internal behavior cannot be easily understood, and for which therefore it is not easy to understand (and verify) how a model has reached a certain conclusion or prediction. The opaqueness of a ML solution may vary depending on the complexity of the underlying model and learning mode.
For example, neural networks tend to be more opaque, due to the intrinsic complexity of the underlying algorithm, than decision trees, the internal functioning of which can be more easily understood by humans. This technical opaqueness is directly linked to the opposing concept of explainability.
A model is explainable when it is possible to generate explanations that allow humans to understand (i) how a result is reached or (ii) on what grounds the result is based (similar to a justification).
Explainability helps business leaders understand why a company is doing what they’re doing with AI. “explainability.”
That means sorting out what an AI algorithm did, what data was used, and why certain conclusions were reached. If, say, a machine learning (ML) algorithm also made business decisions, these decisions need to be annotated and presented effectively. it will be incumbent on AI specialists to show that their data is free of bias and that the outcomes their programs reach are consistent—an interesting challenge for things like deep learning, where there are many, many layers of analysis and different approaches that can affect the outcome.
With algorithms being more and more involved in decision-making processes that can have a significant impact on the lives of those concerned, it is important to understand how they “reason”. as a society we cannot allow certain important decisions to be made with no explanation: “Without being able to explain decisions taken by autonomous systems, it is difficult to justify them: it would seem inconceivable to accept what cannot be justified in areas as crucial to the life of an individual as access to credit, employment, accommodation, justice and health”.
Algorithms can take “bad” decisions due to errors or biases of human origin ( mostly deliberate in Israel against Palestinians ) that are present in the datasets or the code. By making their reasoning transparent, explainability helps to identify the source of these errors and biases, and to correct them. This question is pivotal to the future of AI, as a lack of public confidence could hinder its development.
Explainability – or interpretability – is a component of algorithm transparency. It describes the AI system’s property of being easily understandable by humans. The information must therefore be presented in a form that is intelligible for experts (programmers, data scientists, researchers, etc.) but also for the general public.
Publishing the source code is not enough, not only because that doesn’t systematically make it possible to identify algorithmic bias (the running of certain algorithms cannot be apprehended independently from the training data), but also because it is not readable by a large majority of the public.
Furthermore, this could be in conflict with intellectual property rights, as an algorithm’s source code can be assimilated with a trade secret.
What’s more, X-AI holds several challenges. The first is in the complexity of certain algorithms, based on machine learning techniques such as deep neural networks or random forests, which are intrinsically difficult to grasp for humans; then there is the large quantity of variables that are taken into account.
Second challenge: it’s precisely this complexity that has made algorithms more efficient. In the current state of the art, increasing explainability is often achieved at the expense of precision of the results.
Algorithms that explain algorithms.. Algorithm refers to a set of rules/instructions that step-by-step define how a work is to be executed upon in order to get the expected results.
In order for some instructions to be an algorithm, it must have the following characteristics:
Clear and Unambiguous: Algorithm should be clear and unambiguous. Each of its steps should be clear in all aspects and must lead to only one meaning.
Well-Defined Inputs: If an algorithm says to take inputs, it should be well-defined inputs.
Well-Defined Outputs: The algorithm must clearly define what output will be yielded and it should be well-defined as well.
Finite-ness: The algorithm must be finite, i.e. it should not end up in an infinite loops or similar.
Feasible: The algorithm must be simple, generic and practical, such that it can be executed upon will the available resources. It must not contain some future technology, or anything.
Language Independent: The Algorithm designed must be language-independent, i.e. it must be just plain instructions that can be implemented in any language, and yet the output will be same, as expected.
Inorder to write an algorithm, following things are needed as a pre-requisite:--
The problem that is to be solved by this algorithm.
The constraints of the problem that must be considered while solving the problem.
The input to be taken to solve the problem.
The output to be expected when the problem the is solved.
The solution to this problem, in the given constraints.
Then the algorithm is written with the help of above parameters such that it solves the problem.
To be directly understandable, the algorithm should therefore have a low level of complexity and the model should be relatively simple.
Transparency consists therefore in making data, features, algorithms and training methods available for external inspection and constitutes a basis for building trustworthy models.
Explainable AI is the set of capabilities that describes a model, highlights its strengths and weaknesses, predicts its likely behavior, and identifies any potential biases. It has the ability to articulate the decisions of a descriptive, predictive or prescriptive model to enable accuracy, fairness, accountability, stability and transparency in algorithmic decision making.
Visualization approaches for seeing and understanding the data in the context of training and interpreting machine learning algorithms.
Algorithms: Such as spell check or phonetic algorithms can be useful – but they can also make the wrong suggestion.
AI initiatives are rendered useless, and in some cases detrimental, without clean data to feed their algorithms.
The ability to explain a model’s behavior, answering to an ML engineer, "why did the model predict that?" For example, the prior on variable alpha must not be Gaussian, as we can see in the misaligned posterior predictive check
The ability to translate a model to business objectives, answering in natural language, "why did the model predict that?" For example, the predicted spike in insulin levels are correlated to the recent prolonged inactivity picked up from the fitness watch.
Both definitions are clearly useful. The low-level notion of interpretability lends itself to the engineer's ability to develop and debug models and algorithms. High-level transparency and explainability is just as necessary, for humans to understand and trust predictions in areas like financial markets and medicine
No matter the definition, developing an AI system to be interpretable is typically challenging and ambiguous. It is often the case that a model or algorithm is too complex to understand or describe because its purpose is to model a complex hypothesis or navigate a high-dimensional space, a catch-22. Not to mention what is interpretable in one application may be useless in another.
Even with improved methods and algorithms for explaining AI models and predictions, two core issues must first be addressed in order to make legitimate progress towards interpretable, transparent AI: underspecification and misalignment.
The notion of model or algorithmic interpretability is underspecified -- that is, the AI field is without precise metrics of interpretability. How can one argue a given model or algorithm is more interpretable than another, or benchmark improvements in explainability?
One method could provide beautifully detailed visualizations, while the other provides coherent natural language rationale behind each prediction. It can be apples-and-oranges to compare models on account of their interpretability.
Simple exhaustive searches are rarely sufficient for most real-world problems: the search space (the number of places to search) quickly grows to astronomical numbers. The result is a search that is too slow or never completes. The solution, for many problems, is to use "heuristics" or "rules of thumb" that prioritize choices in favor of those that are more likely to reach a goal and to do so in a shorter number of steps.
In some search methodologies heuristics can also serve to entirely eliminate some choices that are unlikely to lead to a goal (called "pruning the search tree"). Heuristics supply the program with a "best guess" for the path on which the solution lies. Heuristics limit the search for solutions into a smaller sample size..
In computer science, artificial intelligence, and mathematical optimization, a heuristic is a technique designed for solving a problem more quickly when classic methods are too slow, or for finding an approximate solution when classic methods fail to find any exact solution. In computing, heuristic refers to a problem-solving method executed through learning-based techniques and experience.
When exhaustive search methods are impractical, heuristic methods are used to find efficient solutions. A heuristic algorithm is one that is designed to solve a problem in a faster and more efficient fashion than traditional methods by sacrificing optimality, accuracy, precision, or completeness for speed.
Heuristic algorithms often times used to solve NP-complete problems, a class of decision problems. In these problems, there is no known efficient way to find a solution quickly and accurately although solutions can be verified when given. Heuristics can produce a solution individually or be used to provide a good baseline and are supplemented with optimization algorithms.
Heuristic algorithms are most often employed when approximate solutions are sufficient and exact solutions are necessarily computationally expensive A Heuristic is a technique to solve a problem faster than classic methods, or to find an approximate solution when classic methods cannot.
This is a kind of a shortcut as we often trade one of optimality, completeness, accuracy, or precision for speed. A Heuristic (or a heuristic function) takes a look at search algorithms. At each branching step, it evaluates the available information and makes a decision on which branch to follow. It does so by ranking alternatives.
The Heuristic is any device that is often effective but will not guarantee work in every case. three heuristics—availability, representativeness, and anchoring and adjustment. Subsequent work has identified many more. Heuristics that underlie judgment are called "judgment heuristics".
Heuristics can be mental shortcuts that ease the cognitive load of making a decision. Examples that employ heuristics include using a rule of thumb, an educated guess, an intuitive judgment, a guesstimate, profiling, or common sense.
A heuristic technique, often called simply a heuristic, is any approach to problem solving, learning, or discovery that employs a practical method not guaranteed to be optimal or perfect, but sufficient for the immediate goals Heuristics are simple strategies to form judgments and make decisions by focusing on the most relevant aspects of a complex problem. As far as we know, animals have always relied on heuristics to solve adaptive problems, and so have humans
A heuristic function, also called simply a heuristic, is a function that ranks alternatives in search algorithms at each branching step based on available information to decide which branch to follow. For example, it may approximate the exact solution.
Heuristic search refers to a search strategy that attempts to optimize a problem by iteratively improving the solution based on a given heuristic function or a cost measure. Heuristic search is an AI search technique that employs heuristic for its moves. Heuristic is a rule of thumb that probably leads to a solution. ...
Heuristics help to reduce the number of alternatives from an exponential number to a polynomial number. A heuristic function, or simply a heuristic, is a function that ranks alternatives in search algorithms at each branching step based on available information to decide which branch to follow
Examples that employ heuristics include using a rule of thumb, an educated guess, an intuitive judgment, a guesstimate, profiling, or common sense.. heuristic describes a rule or a method that comes from experience and helps you think through things, like the process of elimination, or the process of trial and error.
You can think of a heuristic as a shortcut. Examples of heuristics include using: A rule of thumb. An educated guess The simplest way to describe them is as follows: A heuristic is a rule, strategy or similar mental shortcut that one can use to derive a solution to a problem. A heuristic that works all of the time is known as an algorithm. ...
A systematic error that results from the use of a heuristic is called a cognitive bias The heuristic question is the simpler question that you answer instead. ... On some occasions , substitution will occur and a heuristic answer will be endorsed by System 2. Of course, System 2 has the opportunity to reject this intuitive answer, or to modify it by incorporating other information Heuristic inquiry involves exploring the subjective experience of a particular phenomenon within a purposive sample of individuals.
Heuristic research- ers do not separate the individual from the experience but rather focus their exploration on the essential nature of the relationship or interaction between both. The Affect Heuristic and Decision Making. The affect heuristic is a type of mental shortcut in which people make decisions that are heavily influenced by their current emotions. Essentially, your affect (a psychological term for emotional response) plays a critical role in the choices and decisions you make
Generally speaking, a heuristic is a "rule of thumb," or a good guide to follow when making decisions. In computer science, a heuristic has a similar meaning, but refers specifically to algorithms. ... As more sample data is tested, it becomes easier to create an efficient algorithm to process similar types of data
“Heuristic”
It is based on the psychological principles of "trial and error" theory. Bias occurs when you interpret subsequent information around the anchor. A heuristic is a problem solving approach used to accelerate the process of finding a satisfactory solution. It is a mental shortcut that eases cognitive load when making decisions.
It is a good guess often not made with strong reasoning The heuristic function is a way to inform the search about the direction to a goal. It provides an informed way to guess which neighbor of a node will lead to a goal. There is nothing magical about a heuristic function. It must use only information that can be readily obtained about a node.
When our heuristics fail to produce a correct judgment, it can sometimes result in a cognitive bias, which is the tendency to draw an incorrect conclusion in a certain circumstance based on cognitive factors. ...
This mismatch between our judgment and reality is the result of a bias “A heuristic technique, often called simply a heuristic, is any approach to problem solving, learning, or discovery that employs a practical method not guaranteed to be optimal or perfect, but sufficient for the immediate goals
The heuristic-systematic model is a theory of persuasion that suggests attitudes can change in two fundamentally different ways. ... This simplified form of attitude judgment is called heuristic processing, and it involves using rules of thumb known as heuristics to decide what one's attitudes should be.
The accuracy-effort trade-off theory states that humans and animals use heuristics because processing every piece of information that comes into the brain takes time and effort. With heuristics, the brain can make faster and more efficient decisions, albeit at the cost of accuracy
It is a thorough assessment of a product's user interface, and its purpose is to detect usability issues that may occur when users interact with a product, and identify ways to resolve them. The heuristic evaluation process is conducted against a predetermined set of usability principles known as heuristics.
The classic example of heuristic search methods is the travelling salesman problem. generate a possible solution which can either be a point in the problem space or a path from the initial state. test to see if this possible solution is a real solution by comparing the state reached with the set of goal states.
Heuristics that underlie judgment are called "judgment heuristics". Heuristic Function is a function that estimates the cost of getting from one place to another (from the current state to the goal state.)
Also called as simply a heuristic. Used in a decision process to try to make the best choice of a list of possibilities (to choose the move more likely to lead to the goal state.) A heuristic is a mental shortcut that allows people to solve problems and make judgments quickly and efficiently.
These rule-of-thumb strategies shorten decision-making time and allow people to function without constantly stopping to think about their next course of action It sounds fancy, but you might know a heuristic as a "rule of thumb."
Derived from a Greek word that means "to discover," heuristic describes a rule or a method that comes from experience and helps you think through things, like the process of elimination, or the process of trial and error Heuristic method is a pure discovery method of learning science independent of teacher. ...
In this the teacher set a problem for the students and then stands aside while discover the answer . 2. The method requires the students to solve a number of problems experimentally
AN EXAMPLE OF WHERE HEURISTICS GOES WRONG IS WHENEVER YOU BELIEVE THAT CORRELATION IMPLIES CAUSATION.
Correlation is a relationship or connection between two variables where whenever one changes, the other is likely to also change. But a change in one variable doesn’t cause the other to change. That’s a correlation, but it’s not causation.
Spurious correlations--It is a mathematical relationship in which two or more events or variables are associated but not causally related, due to either coincidence or the presence of a certain third, unseen factor
To find causation, we need explainability. In the era of artificial intelligence and big data analysis, this topic becomes increasingly more important. AIs make data-based recommendations. Sometimes, humans can’t see any reason for those recommendations except that an AI made them. In other words, they lack explainability.
Correlation is about analyzing static historical datasets and considering the correlations that might exist between observations and outcomes. However, predictions don’t change a system. That’s decision making. To make software development decisions, we need to understand the difference it would make in how a system evolves if you take an action or don’t take action. Decision making requires a casual understanding of the impact of an action.
AI technology can’t take its previous learnings from one context and apply them to another situation. Sure, it can identify correlations. But it has no idea which one caused the other, or if that’s even the case.
“Too much of deep learning has focused on correlation without causation, and that often leaves deep learning systems at a loss when they are tested on conditions that aren’t quite the same as the ones they were trained on Giving AI the ability to understand causality would unlock a second renaissance for the technology
Most machine learning-based data science focuses on predicting outcomes, not understanding causality Current approaches to machine learning assume that the trained AI system will be applied on the same kind of data as the training data. In real life it is often not the case.
When humans rationalize the world, we often think in terms of cause and effect — if we understand why something happened, we can change our behavior to improve future outcomes. Causal inference is a statistical tool that enables our AI and machine learning algorithms to reason in similar ways.
Let’s say we’re looking at data from a network of servers. We’re interested in understanding how changes in our network settings affect latency, so we use causal inference to proactively choose our settings based on this knowledge.
The future of AI depends on building systems with notions of causality. while machine learning methods excel in describing the real world, they’re often lacking in understanding the world -- simple perturbations hardly noticed by humans can cause state-of-art deep learning systems to misclassify road signs… The formal modeling and logic to support seemingly fundamental causal reasoning has been lacking in data science and AI
deep learning, and most machine learning (ML) methods for that matter, learn patterns or associations from data. On its own, observational data can only possibly convey associations between variables -- the familiar adage correlation does not imply causation
In correlated data, a pair of variables are related in that one thing is likely to change when the other does. This relationship might lead us to assume that a change to one thing causes the change in the other.
The human brain simplifies incoming information, so we can make sense of it. Our brains often do that by making assumptions about things based on slight relationships, or bias. But that thinking process isn’t foolproof. An example is when we mistake correlation for causation. Bias can make us conclude that one thing must cause another if both change in the same way at the same time
Correlation is a relationship or connection between two variables where whenever one changes, the other is likely to also change. But a change in one variable doesn’t cause the other to change. That’s a correlation, but it’s not causation. There are many forms of cognitive bias or irrational thinking patterns that often lead to faulty conclusions and economic decisions.
These types of cognitive bias are some reasons why people assume false causations in business and marketing: Putting too much weight on your own personal beliefs, over-confidence, and other unproven sources of information often produce an illusion of casualty.
It’s easy to watch correlated data change in tandem and assume that one thing causes the other. That’s because our brains are wired for cause-relation cognitive bias. We need to make sense of large amounts of incoming data, so our brain simplifies it. This process is called heuristics, and it’s often useful and accurate. But not always.
An example of where heuristics goes wrong is whenever you believe that correlation implies causation.
It is a mathematical relationship in which two or more events or variables are associated but not causally related, due to either coincidence or the presence of a certain third, unseen factor To find causation, we need explainability.
In the era of artificial intelligence and big data analysis, this topic becomes increasingly more important. AIs make data-based recommendations. Sometimes, humans can’t see any reason for those recommendations except that an AI made them. In other words, they lack explainability.
Causation takes a step further than correlation. It says any change in the value of one variable will cause a change in the value of another variable, which means one variable makes other to happen. It is also referred as cause and effect Causation, also known as cause and effect, is when an observed event or action appears to have caused a second event or action.
False Causality. To falsely assume when two events occur together that one must have caused the other. While causation and correlation can exist at the same time, correlation does not imply causation.
Causation explicitly applies to cases where action A Causation explicitly applies to cases where action A causes outcome B.causes outcome B. On the other hand, correlation is simply a relationship. Action A relates to Action B—but one event doesn’t necessarily cause the other event to happen.
A heuristic technique, often called simply a heuristic, is any approach to problem solving, learning, or discovery that employs a practical method not guaranteed to be optimal or perfect, but sufficient for the immediate goals Heuristics can be mental shortcuts that ease the cognitive load of making a decision.
Examples that employ heuristics include using trial and error, a rule of thumb, an educated guess, an intuitive judgment, a guesstimate, profiling, or common sense Where finding an optimal solution is impossible or impractical, heuristic methods can be used to speed up the process of finding a satisfactory solution. Heuristics can be mental shortcuts that ease the cognitive load of making a decision.
Examples that employ heuristics include using trial and error, a rule of thumb, an educated guess, an intuitive judgment, a guesstimate, profiling, or common sense. Heuristics are simple strategies or mental processes that humans, animals, organizations and even some machines use to quickly form judgments, make decisions, and find solutions to complex problems. This happens when an individual, human or otherwise, focuses on the most relevant aspects of a problem or situation to formulate a solution.
Those involved in making these decisions can also be influenced by similar past experiences as well. This is the reason that people do not generally stress test every chair or surface they might choose to sit on. Heuristic processes can easily be confused with the use of human logic, and probability. While these processes share some characteristics with heuristics, the assertion that heuristics are not as accurate as logic and probability misses the crucial distinction between risk and uncertainty.
Risk refers to situations where all possible outcomes of an action are known and taken into account when making a decision. In contrast, uncertainty refers to situations where pieces of information are unknown or unknowable In situations of risk, heuristics face an accuracy-effort trade-off where their simplified decision process leads to reduced accuracy. In contrast, situations of uncertainty allow for less-is-more effects, where systematically ignoring (or in some cases lacking) information leads to more accurate inferences.
Less-is-more effects have been shown experimentally, analytically, and by computer simulations. Though both of these mental processes are similar to heuristics, they are not the same. heuristics are concerned with finding a solution that is "good enough" to satisfy a need. They serve as a quick mental reference for everyday experiences and decisions.
Understanding cause and effect would make existing AI systems smarter and more efficient. A robot that understands that dropping things causes them to break would not need to toss dozens of vases onto the floor to see what happens to them. “Humans don't need to live through many examples of accidents to drive prudently,. They can just imagine accidents, “in order to prepare mentally if it did actually happen.”
Algorithms are not the computer code. They are just the instructions which give a clear idea to write the computer code.
Qualities of a good algorithm:---
Input and output should be defined precisely.
Each step in the algorithm should be clear and unambiguous.
Algorithms should be most effective among many different ways to solve a problem.
An algorithm shouldn't have computer code. Instead, the algorithm should be written in such a way that, it can be used in different programming languages.
An algorithm is a series of steps for solving a problem, completing a task or performing a calculation. Algorithms are usually executed by computer programs .. Algorithm is a precise step-by-step plan for a computational procedure that possibly begins with an input value and yields an output value in a finite number of steps while code is a short symbol, often with little relation to the item it represents.
An algorithm is an idea, a process, a recipe, etc. It's a sequence of steps, a procedure, that can be used to produce a result. It is independent of any programming language. Code is practical realization of algorithm.
An algorithm is an idea, a process, a recipe, etc. It's a sequence of steps, a procedure, that can be used to produce a result. It is independent of any programming language. ... When you then implement that algorithm by coding it in some language, that is code. Algorithms can be expressed using natural language, flowcharts, etc.
As a job title, “programmer”, “software developer”, and “software engineer” can mean whatever a given company wants them to mean. Some places call the people who create software “engineers”, others call them “developers”, and still others call them “programmers”. A programmer programs. A software developer develops software.
A software engineer engineers software systems. They’re 3 different hats that the same people often wear at different times. In the US the word “engineer” often has an actual legal meaning with licensing requirements one is expected to meet before applying the term to oneself.
In order to become an actual licensed software engineer, you would have to get a degree in software engineering.. someone who designs and creates a program from scratch
Autonomous and adaptive analytics: this technique is the most complex and uses forward looking predictive analytics models that automatically learn from transactions and update results in real time using ML. This includes the ability to self-generate new algorithmic models with suggested insights for future tasks, based on correlations and patterns in the data that the system has identified and on growing volumes of Big Data.
The Adaptive Learning process monitors and learns the new changes made to the input and output values and their associated characteristics. In addition, it learns from the events that may alter the market behavior in real time and, hence, maintains its accuracy at all times. Adaptive AI accepts the feedback received from the operating environment and acts on it to make data-informed predictions.
Advanced analytics often uses ML to gain deeper insights, make predictions or generate recommendations for business purposes. This is done by means of suitable ML algorithms able to recognise common patterns in large volumes of data via a learning (or ‘training’) process. The result of the learning process is a model, which represents what the algorithm has learnt from the training data and which can be used to make predictions based on new input data
In machine learning, a common task is the study and construction of algorithms that can learn from and make predictions on data. Such algorithms work by making data-driven predictions or decisions, through building a mathematical model from input data. The data used to build the final model usually comes from multiple datasets.
Model training (also called ‘learning’) consists in feeding the training dataset to the algorithm tobuild the model. The challenge of this phase is to build a model that fits the given dataset with sufficient accuracy and has a good generalisation capability on unseen data, i.e. a model that is fit.
An ML model generalises well when its predictions on unseen observations are of a similar quality (accuracy) to those made on test data.
After the training phase, models are calibrated/ tuned by adjusting their hyper-parameters. Examples of hyper-parameters are the depth of the tree in a decision tree algorithm, the number of trees in a random forest algorithm, the number of clusters k in a k-means algorithm, the number of layers in a neural network, etc.
Selection of incorrect hyper-parameters can result in the failure of the model.
Data mining of Big Data is achieved by using AI programming that works with algorithms to find patterns in the Big Data that are noteworthy. This provides insights that help management make better-informed decisions.
Upwrds of 70% of the time and energy spent in AI projects is consumed by preparing the data to be consumed by the ML algorithms. This data-management work can be broken down into a handful of phases, including: ---
Data discovery: What data do you need, and how do you find it?
Data analysis: Is the data useful for ML? How do you validate it?
Data preparation: What format is the data in? How do you transform it into the correct format?
Data modeling: What algorithm can you use to model the data?
Model evaluation: Is the model working as expected? What changes do you need to make?
Before an algorithm can train on a piece of data, it must be converted into a machine-readable format that it understands, which is another critical step in the AI and ML process. Scientists may have to encode the data a certain way, or use bucketization or binning techniques to represent data values in certain ranges.
Once the data is fully prepped and in the correct format, the data scientist can write an algorithm, or a pre-written one, to model the data. This starts the productionalization phase of the AI and ML journey, which has its own set of challenges.
if we want text analysis software to perform desired tasks, we need to teach machine learning algorithms how to analyze, understand and derive meaning from text. But how? The simple answer is by tagging examples of text. Once a machine has enough examples of tagged text to work with, algorithms are able to start differentiating and making associations between pieces of text, and can even begin to make predictions.
Clustering
Text clusters are able to understand and group vast quantities of unstructured data. Although less accurate than classification algorithms, clustering algorithms are faster to implement because you don't need to tag examples to train models. That means these smart algorithms mine information and make predictions without the use of training data, otherwise known as unsupervised machine learning.
Google is a great example of how clustering works. When you search for a term on Google, have you ever wondered how it takes just seconds to pull up relevant results? Google's algorithm breaks down unstructured data from web pages and groups pages into clusters around a set of similar words or n-grams (all possible combinations of adjacent words or letters in a text). So, the pages from the cluster that contains a higher count of words or n-grams relevant to the search query will appear first within the results.
Consistent Criteria
Humans make errors. Fact. And the more tedious and time-consuming a task is, the more errors that are made. By using automated text analysis models that have been trained, algorithms are able to analyze, understand, and sort through data more accurately than humans.
We are influenced by personal experiences, thoughts, and beliefs when reading texts, whereas algorithms are influenced by the information they've received. By applying the same criteria to analyze all data, algorithms are able to deliver more consistent and reliable data.
AI works by combining large amounts of data with fast, iterative processing and intelligent algorithms, allowing the software to learn automatically from patterns or features in the data.
The naive Bayes algorithm uses a slight modification of the Bayes formula to determine the probability that certain words belong to text of a specific type. The ‘naive’ part of naive Bayes come from the fact the algorithm treats each word independently and a text is considered simply as a set of words. The wider context of the words is then lost using naive Bayes.
Algorithms are being developed to help pilot cars, guide weapons, perform tedious or dangerous work, engage in conversations, recommend products, improve collaboration, and make consequential decisions in areas such as jurisprudence, lending, medicine, university admissions, and hiring.
But while the technologies enabling AI have been rapidly advancing, the societal impacts are only beginning to be fathomed if automatic techniques or naïve statistical methodologies are used to train algorithms on data that contain inaccuracies or biases, those algorithms themselves might well reflect those inaccuracies or biases.
A machine learning algorithm is only safe and reliable to the extent that it is trained on (1) sufficient volumes of data that (2) are suitably representative of the scenarios in which the algorithm is to be deployed.
Artificial intelligence is leading us toward a new algorithmic warfare battlefield that has no boundaries or borders, may or may not have humans involved, and will be impossible to understand and perhaps control across the human ecosystem in cyberspace, geospace, and space (CGS).
As a result, the very idea of the weaponization of artificial intelligence, where a weapon system that, once activated across CGS, can select and engage human and non-human targets without further intervention by a human designer or operator, is causing great concern.
The thought of an intelligent machine or machine intelligence to have the ability to perform any projected warfare task without any human involvement and intervention -- using only the interaction of its embedded sensors, computer programming, and algorithms in the human environment and ecosystem -- is becoming a reality that cannot be ignored anymore.
The rapid development of AI weaponization is evident across the board: navigating and utilizing unmanned naval, aerial, and terrain vehicles, producing collateral-damage estimations, deploying “fire-and-forget” missile systems and using stationary systems to automate everything from personnel systems and equipment maintenance to the deployment of surveillance drones, robots and more are all examples.
algorithms are by no means secure—nor are they immune to bugs, malware, bias, and manipulation. And, since machine learning uses machines to train other machines, what happens if there is malware or manipulation of the training data?
While security risks are everywhere, connected devices increase the ability of cybersecurity breaches from remote locations and because the code is opaque, security is very complex. So, when AI goes to war with other AI (irrespective if that is for cyber-security, geo-security, or space-security), the ongoing cybersecurity challenges will add monumental risks to the future of humanity and the human ecosystem in CGS.
There are weapons that use artificial intelligence in active use today, including some that can search, select and engage targets on their own, attributes often associated with defining what constitutes a lethal autonomous weapon system (a.k.a. a killer drone/ robot).
the Israel Aerospace Industries Harpy, an armed drone that can hang out high in the skies surveying large areas of land until it detects an enemy radar signal, at which point it crashes into the source of the radar, destroying both itself and the target.
The weapon needs no specific target to be launched, and a human is not necessary to its lethal decision making. spooky strand of research seeks to build algorithms that tip human analysts off to such targets by singling out cars driving suspiciously around a surveilled city.
An actor with darker motives might use algorithms as a convenient veil for an intentionally insidious decisions.
Automation’s vast potential to make humans more efficient extends to the very human act of committing war crimes.
If one system offers up a faulty conclusion, it could be easy to catch the mistake before it does any harm. But these algorithms won’t act alone. A few months ago, the U.S. Navy tested a network of three AI systems, mounted on a satellite and two different airplanes, that collaboratively found an enemy ship and decided which vessel in the Navy’s fleet was best placed to destroy it, as well as what missile it should use. The one human involved in this kill chain was a commanding officer on the chosen destroyer, whose only job was to give the order to fire.
Eventually, the lead-up to a strike may involve dozens or hundreds of separate algorithms, each with a different job, passing findings not just to human overseers but also from machine to machine. Mistakes could accrue; human judgment and machine estimations would be impossible to parse from one another; and the results could be wildly unpredictable.
Militaries have long argued that AI will make conflict more precise. But that argument has a dark flipside: An algorithm designed to minimize civilian casualties could just as easily be used to calculate how civilian harm could be maximized.
Governments must develop inscrutable and transparent mechanisms to audit algorithms that go bad, as well as those humans who employ their algorithms badly.
The drones are required to communicate with each other and even respond to each other. Therefore they require sensors and they require decision-making based algorithms – and thus this is the ‘AI and autonomy’ stage.AI Worldwide for Warfare
Unlike human intelligence, AI algorithms do not possess common sense, conceptual understanding, notions of cause-and-effect, or intuitive physics.
Their lack of common sense, the inability to generalize or to consider context, makes AI algorithms “brittle,” meaning that they cannot handle unexpected scenarios or unfamiliar situations. As causes that may lead to unfairness in machine learning :---
• Biases already included in the datasets used for learning, which are based on biased device measurements, historically biased human decisions, erroneous reports or other reasons. Machine learning algorithms are essentially designed to replicate these biases.
• Biases caused by missing data, such as missing values or sample/selection biases, which result in datasets that are not representative of the target population.
• Biases that stem from algorithmic objectives, which aim at minimizing overall aggregated prediction errors and therefore benefit majority groups over minorities.
• Biases caused by "proxy" attributes for sensitive attributes. Sensitive attributes differentiate privileged and unprivileged groups, such as race, gender and age, and are typically not legitimate for use in decision making. Proxy attributes are non-sensitive attributes that can be exploited to derive sensitive attributes. In the case that the dataset contains proxy attributes, the machine learning algorithm can implicitly make decisions based on the sensitive attributes under the cover of using presumably legitimate attributes .
Algorithms make predictions that mirror past patterns. This new data is then fed back into the technological model, creating a pernicious feedback loop in which social injustice is not only replicated, but in fact further entrenched.
It is also worth noting that the same communities that have been overpoliced have been severely neglected, both intentionally and unintentionally, in many other areas of social and political life. While they are overrepresented in crime rate data sets, they are underrepresented in many other data sets
CJI BOBDE WANTS TO USE AI IN INDIAN JUDICIARY.. WARNING, NEVER GIVE A COMPUTER TO A MONKEY
Today 1 in 33 Americans are under some form of correctional supervision.
Risk assessment tools are designed to do one thing: take in the details of a defendant’s profile and spit out a recidivism score—a single number estimating the likelihood that he or she will reoffend. A judge then factors that score into a myriad of decisions that can determine what type of rehabilitation services particular defendants should receive, whether they should be held in jail before trial, and how severe their sentences should be. A low score paves the way for a kinder fate. A high score does precisely the opposite.
RISK ASSESSMENT CAN NEVER BE OBJECTIVE . SUBJECTIVE ( CONSCIOUS HUMAN ) MUST HAVE VETO POWER
The logic for using such algorithmic tools is that if you can accurately predict criminal behavior, you can allocate resources accordingly, whether for rehabilitation or for prison sentences. Judges are making STUPID OBJECTIVE decisions on the basis of data-driven recommendations and not SUBJECTIVE DECISIONS based on wisdom. .
Machine-learning algorithms use statistics to find patterns in data. So if you feed it historical crime data, it will pick out the patterns associated with crime. But those patterns are statistical correlations—nowhere near the same as causations.
If an algorithm found, for example, that low income was correlated with high recidivism, it would leave you none the wiser about whether low income actually caused crime. But this is precisely what risk assessment tools do: they turn correlative insights into causal scoring mechanisms.
Now populations that have historically been disproportionately targeted by law enforcement—especially low-income and minority communities—are at risk of being slapped with high recidivism scores like Palestinians in Israel.
The algorithm could amplify and perpetuate embedded biases and generate even more bias-tainted data to feed a vicious cycle. Because most risk assessment algorithms are proprietary, it’s also impossible to interrogate their decisions or hold them accountable.
Humans have empathy built in because we evolved to be social animals. An artificial intelligence built from the ground needn't come with empathy. If we don't make sure to build empathy into such AI at the onset it could be dangerous for us
WE WILL NOT ALLOW STARE DECISIS TO BE SUSTAINED BY AI. STARE DECISIS IS ILLEGAL AND HAS BEEN EXPLOITED BY TRAITOR JUDGES IN FOREIGN PAYROLL TO CREATE THE NAXAL RED CORRIDOR AND CAUSE ETHNIC CLEANSING OF KASHMIRI PANDITS..
WHEN IT SUITS THE AGENDA OF THESE TRAITOR FOREIGN PAYROLL JUDGES THEY DECLARE THAT MAJORITARIANISM HARMS DEMOCARCY..
ARTIFICIAL INTELLIGENCE IS WHEN THE CODE IS SELF AWARE .. ANY IDIOT KNOWS THAT THIS IS IMPOSSIBLE . THE GOAL OF AI IS TO TAKE OVER DECISIONS THAT WE USUALLY TAKE AS HUMANS.
HUMANS HAVE THINGS A COMPUTER CAN NEVER HAVE.. A SUBCONSCIOUS BRAIN LOBE, REM SLEEP WHICH BACKS UP BETWEEN RIGHT/ LEFT BRAIN LOBES AND FROM AAKASHA BANK, A GUT WHICH INTUITS, 30 TRILLION BODY CELLS WHICH HOLD MEMORY, A VAGUS NERVE , AN AMYGDALA , 73% WATER IN BRAIN FOR MEMORY, 10 BILLION MILES ORGANIC DNA MOBIUS WIRING ETC.
HUMANS HAVE THINGS A COMPUTER CAN NEVER HAVE.. A SUBCONSCIOUS BRAIN LOBE, REM SLEEP WHICH BACKS UP BETWEEN RIGHT/ LEFT BRAIN LOBES AND FROM AAKASHA BANK, A GUT WHICH INTUITS, 30 TRILLION BODY CELLS WHICH HOLD MEMORY, A VAGUS NERVE , AN AMYGDALA , 73% WATER IN BRAIN FOR MEMORY, 10 BILLION MILES ORGANIC DNA MOBIUS WIRING ETC.
The software is able to understand itself, and automatically understand and respond differently to different situations. AI today is stupid AI. It's just dumb algorithms that try to do clever stuff
AI CAN BE USED TO HELP OUT STUPID COLLEGIUM JUDICIARY WITH BODMAS
https://timesofindia.indiatimes.com/city/mumbai/mumbai-noida-kerala-raids-blow-lid-off-phone-racket/articleshow/74025539.cms
Complex AI algorithms allow organizations to unlock insights from data that were previously unattainable. However, the blackbox nature of these systems means it isn't straightforward for business users to understand the logic behind the decision. Even the data scientists that created the model may have trouble explaining why their algorithm made a particular decision.
One way to achieve better model transparency is to adopt from a specific family of models that are considered explainable. Examples of these families include linear models, decision trees, rules sets, decision sets, generalized additive models and case-based reasoning methods.
It is useful to distinguish between the concepts of procedural and distributive fairness.
Procedural justice concerns the fairness and the transparency of the processes by which decisions are made, and may be contrasted with distributive justice (fairness in the distribution of rights or resources), and retributive justice (fairness in the punishment of wrongs) Procedural justice is the idea of fairness in the processes that resolve disputes and allocate resources.
One aspect of procedural justice is related to discussions of the administration of justice and legal proceedings. Procedural fairness is concerned with the procedures used by a decision maker, rather than the actual outcome reached. It requires a fair and proper procedure be used when making a decision.
Procedural justice is when employees perceive that the processes that lead to important outcomes are fair and just. For example, the process of how a manager gives raises will be seen as unfair if he only gives raises to his friends..
A policy (or an algorithm) is said to be procedurally fair if it is fair independently of the outcomes it produces.
Procedural fairness is related to the legal concept of due process. A policy (or an algorithm) is said to be distributively fair if it produces fair outcomes.
Most ethicists take a distributive view of justice, whereas a procedure’s fairness rests largely on the outcomes it produces. On the other hand, people often tend toward a more procedural view, in some cases caring more about being treated fairly than the outcomes they experience
AI algorithms often attract criticism for being distributively unfair
Distributive justice concerns the socially just allocation of goods. Often contrasted with just process, which is concerned with the administration of law, distributive justice concentrates on outcomes. This subject has been given considerable attention in philosophy and the social sciences.
Distributive justice theory argues that societies have a duty to individuals in need and that all individuals have a duty to help others in need. Proponents of distributive justice link it to human rights.
Five types of distributive norm are defined --
Equality: Regardless of their inputs, all group members should be given an equal share of the rewards/costs. Equality supports that someone who contributes 20% of the group's resources should receive as much as someone who contributes 60%. .
Equity: Members' outcomes should be based upon their inputs. Therefore, an individual who has invested a large amount of input (e.g. time, money, energy) should receive more from the group than someone who has contributed very little. Members of large groups prefer to base allocations of rewards and costs on equity.
Power: Those with more authority, status, or control over the group should receive less than those in lower level positions.
Need: Those in greatest needs should be provided with resources needed to meet those needs. These individuals should be given more resources than those who already possess them, regardless of their input.
Responsibility: Group members who have the most should share their resources with those who have less.
Substantive fairness means there is a fair or valid reason for the employer to dismiss an employee Employers have the right to expect a certain standard of work and conduct from an employee and in turn, an employee should be protected from arbitrary action.
principles of restorative justice--- Crime causes harm and justice should focus on repairing that harm. The people most affected by the crime should be able to participate in its resolution
Data Science is the study of all types of data- structured or unstructured, to gain business insights. It makes use of various techniques and algorithms that help to collect, store and analyze business data and gain valuable information. It allows business organizations to collect and organize the data. They use this data to analyze trends and present the gained information within the organization.
The professionals who apply all the techniques in data and build models on top of that data are Data Scientists. Data Scientists use various scientific algorithms that help to develop business strategies and make necessary changes and improvements in the business.
A confusion matrix is a table that is often used to describe the performance of a classification model (or “classifier”) on a set of test data for which the true values are known. It allows the visualization of the performance of an algorithm.
It is a matrix where we put the actual values in the columns and the predicted values in the rows. Thus the intersection of rows and columns becomes our metrics. A Confusion matrix is the comparison summary of the predicted results and the actual results in any classification problem use case.
The comparison summary is extremely necessary to determine the performance of the model after it is trained with some training data. There are various components that exist when we create a confusion matrix. The components are mentioned below
Positive(P): The predicted result is Positive (Example: Image is a cat)
Negative(N): the predicted result is Negative (Example: Images is not a cat)
True Positive(TP): Here TP basically indicates the predicted and the actual values is 1(True)
True Negative(TN): Here TN indicates the predicted and the actual value is 0(False)
False Negative(FN): Here FN indicates the predicted value is 0(Negative) and Actual value is 1. Here both values do not match. Hence it is False Negative.
False Positive(FP): Here FP indicates the predicted value is 1(Positive) and the actual value is 0. Here again both values mismatches. Hence it is False Positive.
Accuracy and Components of Confusion Matrix
After the confusion matrix is created and we determine all the components values, it becomes quite easy for us to calculate the accuracy. So, let us have a look at the components to understand this better.
Classification Accuracy
Accuracy-formula-Confusion-Matrix
From the above formula, the sum of TP (True Positive) and the TN (True Negative) are the correct predicted results. Hence in order to calculate the accuracy in percentage, we divide with all the other components
In the field of machine learning and specifically the problem of statistical classification, a confusion matrix, also known as an error matrix, is a specific table layout that allows visualization of the performance of an algorithm, typically a supervised learning one (in unsupervised learning it is usually called a matching matrix).
Each row of the matrix represents the instances in a predicted class while each column represents the instances in an actual class (or vice versa). The name stems from the fact that it makes it easy to see if the system is confusing two classes (i.e. commonly mislabeling one as another).
It is a special kind of contingency table, with two dimensions ("actual" and "predicted"), and identical sets of "classes" in both dimensions (each combination of dimension and class is a variable in the contingency table).
A false positive is an error resulting from a test or algorithm indicating the presence of a condition (for instance, being a fraudster, having a rare disease, or being a terrorist) that does not in fact exist. If the overall population-level base rate is low, then even the most sophisticated algorithms often yield more false positives Than true positives. This is known as the “false positives paradox.”
To illustrate, suppose that each year a country faces only a small handful of commercial airline terrorists threats, and that the best available algorithm homes in on a few hundred suspects out of millions of passengers.
Though tiny relative to the overall population, the great majority of people on this list will be innocent. Furthermore, because no algorithm is perfectly accurate, it is quite possible that this list won’t contain all of the actual terrorists, a type of error called a false negative. The tradeoff is that, expanding the list of suspects to reduce
The likelihood of false negatives will increase the number of false positives—and therefore the risk of harming or treating unfairly still more innocent people. Analogous scenarios involve electing algorithmic thresholds for deciding when to treat people at risk of a disease. There is generally a tradeoff between correctly identifying as many people with the disease as possible versus avoiding potentially risky treatments of healthy people.
In the regulation of algorithms, particularly artificial intelligence and its subfield of machine learning, a right to explanation (or right to an explanation) is a right to be given an explanation for an output of the algorithm. Such rights primarily refer to individual rights to be given an explanation for decisions that significantly affect an individual, particularly legally or financially.
However, today algorithms take a myriad of decisions without consulting humans: they have become the decision makers, and humans have been pushed into an artefact shaped by technology.
When it comes to AI, “explanation” could mean several things: 1) How an algorithm works or how the system functions. 2) The factors or data that resulted in a decision by the algorithm or system that impacted an individual (a data subject)
The algorithm is basically a code developed to carry out a specific process. Its a process or set of rules to be followed in calculations or other problem solving operations, usually by a computer.
Algorithmic trading is heavily used by banks and trading institutions..
Algorithmic trading, encompasses trading systems that are heavily reliant on complex mathematical formulas and high-speed, computer programs to determine trading strategies
It is a trading system that utilizes very advanced mathematical models for making transaction decisions in the financial markets.
Algorithmic Trading is a process to Buy or Sell a security based on some pre-defined set of rules which are backtested on Historical data. These rules can be based on Technical Analysis, charts, indicators or even Stock fundamentals.
In algorithmic trading:--
Inputs: quotes, trades in the stock market, liquidity opportunities
Output: intelligent trading decisions.
Algorithmic trading is a method of executing orders using automated pre-programmed trading instructions accounting for variables such as time, price, and volume . This type of trading was developed to make use of the speed and data processing advantages that computers have over human traders.
Algorithmic trading is the use of computers and computer-based models to initiate trades and match buyers with sellers, hence, it can be used for making markets and also for proprietary trading
Artificial intelligence refers to a class of computer programs designed to solve problems requiring inferential reasoning, decision making based on incomplete or uncertain information, classification, optimization, and perception
On the most inflexible end of the spectrum are AI that make decisions based on preprogrammed rules from which they make inferences or evaluate options . On the most flexible end are modern AI programs that are based on machine-learning algorithms that can learn from data.
Such AI would, in contrast to the rule-based AI, examine countless other chess games and dynamically find patterns that it then uses to make moves — it would come up with its own scoring formula For this sort of AI, there are no pre-programmed rules about how to solve the problem at hand, but rather only rules about how to learn from data.
Many modern machine-learning algorithms share their pedigree with the vast array of statistical inference tools that are employed broadly in the physical and social sciences. They may, for example, use methods that minimize prediction error, adjust weights assigned to various variables, or optimize both in tandem.
For instance, a machine-learning algorithm may be given three pieces of data, such as a person’s height, weight, and age, and then charged with the task of predicting the time in which each person in a dataset can run a mile.
The machine-learning algorithm would look through hundreds or thousands of examples of people with various heights, weights and ages and their mile times to devise a model. One simple way to do so would be to assign some co-efficient or weight to each piece of data to redict the mile time. For example:--
Predicted Mile Time = A x Height + B x Weight + C x Age
The algorithm may continue to adjust A, B and C as it goes through the examples it has been given to look for the values for A, B and C that result in the smallest error — that is, the difference between each person in the training data’s actual mile time and the algorithm’s predicted mile time.
This example is the same framework for a least-squares regression, in which the square of the error of the predicting equation is minimized. Many machine-learning algorithms are directed at a similar task but use more mathematically sophisticated methods to determine weights for each variable or to minimize some defined error or “loss function.
Machine-learning algorithms are often given training sets of data to process. Once the algorithm trains on that data, it is then tested with a new set of data used for validation. The goal of tuning a machine-learning algorithm is to ensure that the trained model will generalize, meaning that it has predictive power when given a test dataset (and ultimately live data).
Machine-learning algorithms commonly (though not necessarily) make predictions through categorization. These “classifiers” are able to, for example, look at millions of credit reports and classify individuals into separate credit risk categories or process images and separate the ones containing faces from the ones that do not.
If A machine-learning algorithm is properly generalizing, it will correctly predict the appropriate classification for a particular data point. One possible reason AI may be a black box to humans is that it relies on machine-learning algorithms that internalize data in ways that are not easily audited or understood by humans a lack of transparency may arise from the complexity of the algorithm’s structure, such as with a deep neural network, which consists of thousands of artificial neurons working together in a diffuse way to solve a problem.
This reason for AI being a black box is referred to as “complexity.” The lack of transparency may arise because the AI is using a machine-learning algorithm that relies on geometric relationships that humans cannot visualize, such as with support vector machines. This reason for AI being a black box is referred to as “dimensionality.”
The deep neural network is based on a mathematical model called the artificial neuron. While originally based on a simplistic model of the neurons in human and animal brains, the artificial neuron is not meant to be a computer-based simulation of a biological neuron. Instead, the goal of the artificial neuron is to achieve the same ability to learn from experience as with the biological neuron.
The ability to connect layers of neural networks has yielded staggering results. What has emerged is the so-called “deep” architecture of artificial neurons, referred to as Deep Neural Networks, where several layers of interconnected neurons are used to progressively find patterns in data or to make logical or relational connections between data points.
Deep networks of artificial neurons have been used to recognize images, even detecting cancer at levels of accuracy equalling that of experienced doctors. No single neuron in these networks encodes a distinct part of the decision-making process.
The thousands or hundreds of thousands of neurons work together to arrive at a decision. A layer or cluster of neurons may encode some feature extracted from the data (e.g., an eye or an arm in a photograph), but often what is encoded will not be intelligible to human beings.
The net result is akin to the way one “knows” how to ride a bike. Although one can explain the process descriptively or even provide detailed steps, that information is unlikely to help someone who has never ridden one before to balance on two wheels. One learns to ride a bike by attempting to do so over and over again and develops an intuitive understanding.
Because a neural network is learning from experience, its decision-making process is likewise intuitive. Its knowledge cannot in most cases be reduced to a set of instructions, nor can one in most cases point to any neuron or group of neurons to determine what the system found interesting or important.
Its power comes from “connectionism,” the notion that a large number of simple computational units can together perform computationally sophisticated tasks. The complexity of the large multi-layered networks of neurons is what gives rise to the Black Box Problem.
Some machine-learning algorithms are opaque to human beings because they arrive at decisions by looking at many variables at once and finding geometric patterns among those variables that humans cannot visualize.
Modern AI systems are built on machine-learning algorithms that are in many cases functionally black boxes to humans. At present, it poses an immediate threat to intent and causation tests that appear in virtually every field of law. These tests, which assess what is foreseeable or the basis for decisions, will be ineffective when applied to black-box AI.
The solution to this problem should not be strict liability or a regulatory framework of granularly defined transparency standards for AI design and use. Both solutions risk stifling innovation and erecting significant barriers to entry for smaller firms.
A sliding scale system is a better approach. It adapts the current regime of causation and intent tests, relaxing their requirements for liability when AI is permitted to operate autonomously or when AI lacks transparency, while preserving traditional intent and causation tests when humans supervise AI or when the AI is transparent.
The definition of interpretable AI isn't exactly black and white. To have a productive conversation it's essential to be clear what model interpretability means to different stakeholders:
Deep learning is fundamentally blind to cause and effect.
Unlike a real doctor, a deep learning algorithm cannot explain why a particular image may suggest disease. This means deep learning must be used cautiously in critical situations. Deep learning’s pattern recognition capabilities have revolutionized technology.
Unlike a real doctor, a deep learning algorithm cannot explain why a particular image may suggest disease. This means deep learning must be used cautiously in critical situations. Deep learning’s pattern recognition capabilities have revolutionized technology.
But if it can’t understand cause and effect, AI will never reach its true potential because it will never come close to replicating human intelligence Machine learning applications involving deep learning are usually trained to accomplish a highly specific task such as recognizing spoken commands or images of human faces.
Since its explosion in popularity in 2012, deep learning’s unparalleled ability to recognize patterns in data has led to some incredibly important uses, like uncovering fraud in financial activity and identifying indications of cancer in x-ray scans.
GO OR GOLANG ( MY ELDER SON USES THIS )
Go is syntactically similar to C, but with memory safety, garbage collection, structural typing, and CSP-style concurrency. Go is a Procedural, functional and concurrent language
Go is ideal for system programming.. Go supports concurrency..
There are two major implementations:--
Google's self-hosting compiler toolchain targeting multiple operating systems, mobile devices, and WebAssembly.
gccgo, a GCC frontend.
A third-party transpiler GopherJS compiles Go to JavaScript for front-end web development.
Go is an open-source programming language developed by Google. It is a statically-typed compiled language. This language support concurrent programming and also allows running multiple processes simultaneously. This is achieved using channels, goroutines, etc. Go has garbage collection, which itself does the memory management and allows the deferred execution of functions.
Here, are important reasons for using Go language:--
It allows you to use static linking to combine all dependency libraries and modules into one single binary file based on the type of the OS and architecture.
Go language performed more efficiently because of CPU scalability and concurrency model.
Go language offers support for multiple libraries and tools, so it does not require any 3rd party library.
It's statically, strongly typed programming language with a great way to handle errors
Here, are important reasons for using Go language:--
It allows you to use static linking to combine all dependency libraries and modules into one single binary file based on the type of the OS and architecture.
Go language performed more efficiently because of CPU scalability and concurrency model.
Go language offers support for multiple libraries and tools, so it does not require any 3rd party library.
It's statically, strongly typed programming language with a great way to handle errors
Here, are cons/drawbacks of using GO language:--
Go is not a generic language
API integration with Go does not have an officially supported Go SDK.
Poor Library Support
Fractured Dependency Management
- REGARDING TORTURE / MURDER OF IB OFFICER ANKIT SHARMA BY AN ILLEGAL IMMIGRANT MUSLIM IMMIGRANT MOB LED BY AAP LEADER TAHIR HUSSAIN..
NOTHING WILL HAPPEN AS TRAITOR JUDICIARY AND MEDIA ARE ON THE SIDE OF THE ILLEGAL IMMIGRANT MUSLIMS..
INDIAN COPS / SECURITY AGENCIES HAVE NO PRIDE OR HONOR.
THIS IS WHY CJI GOGOI WAS ABLE TO TREAT CBI CHIEF LIKE A CLASS DUNCE AND MAKE HIM SIT IN A CORNER OF THE COURT ROOM THE WHOLE DAY..
IN REALITY, IT SHOULD HAVE BEEN THE OTHER WAY AROUND.. CBI DIRECTOR WHO HAS TAKEN THE OATH IS NOT SMALL FRY..
NOW-- LET US COMPARE INDIA WITH USA.
MIGUEL ÁNGEL FÉLIX GALLARDO WAS A MEXICAN COCAINE DRUG LORD WHO RAN GUNS (FOR PRESIDENT RONALD REAGAN AND CIA DIRECTOR GEORGE HERBERT WALKER BUSH SENIOR ) TO THE CONTRAS IN NICARAGUA .
MERCENARY CONTRAS WERE CREATED/ ARMED/ FUNDED BY CIA TO TOPPLE THE PATRIOT SANDINISTA GOVT OF NICARAGUA WHO KICKED OUT JEWISH OLIGARCHS WHO WERE LOOTING THE NATION..
PATRIOT SANDINISTAS WERE DUBBED AS BAAAAD COMMIES BY PRESIDENT REAGAN.
US PRESIDENT AND CIA WERE UNDERCUTTING AMERICAN DEA DEPT.. THEY GOT HUGE BRIBES FROM COCAINE DRUG LORD FELIX BYPASSING OFFICIAL PROTOCOL OF US CONGRESS SANCTION OF FUNDS..
BUT DRUG LORD MIGUEL ÁNGEL FÉLIX GALLARDO MADE A BIG MISTAKE..
HE ORDERED THE KILLING OF US DEA AGENT KIKI CAMARENA ( WITH BLESSINGS OF BUSH/ REAGAN ) WHO EXHUMED THE “GUNS FOR COCAINE” CONSPIRACY…
AS SOON AS THIS HAPPENED DEA WENT AGAINST THE WHITE HOUSE AND CIA.. THEY CREATED THEIR OWN UNOFFICIAL ROGUE HIT SQUAD TO TAKE REVENGE ..
https://en.wikipedia.org/wiki/Kiki_Camarena
DEA FOUND OUT DIRECT CIA / WHITE HOUSE INVOLVEMENT IN THE TORTURE AND MURDER OF DEA AGENT KIKI CAMARENA ..
THE ROGUE DEA SQUAD TORTURED AND KILLED WHOEVER WERE INVOLVED IN THE TORTURE OF KIKI CAMARENA ..
DEA EXTRACTED CONFESSIONS FROM A MEXICAN DOCTOR AND A MEXICAN POLICE OFFICER AFTER TORTURING THEM.
A US CIA OFFICER FELIX RODRIGUEZ HAD OVERSEEN THE ENTIRE TORTURE AND KILLING OF THE DEA AGENT ON ORDERS FROM BUSH SR AND REAGAN..
FELIX RODRIGUEZ RAN THE CONTRA SUPPLY DEPOT .. DEAD MEN TELL NO TALES..
REAGAN AND BUSH SR ORDERED THE ARREST OF DRUG LORD MIGUEL ÁNGEL FÉLIX GALLARDO TO SHUT DOWN THIS CASE BEFORE SHIT HIT THE FAN.
https://en.wikipedia.org/wiki/Miguel_%C3%81ngel_F%C3%A9lix_Gallardo
IN USA , IF YOU KILL A COP HIS MATES GO ROGUE TAKE REVENGE .. AND THIS IS UNOFFICIALLY ALLOWED.. THIS IS WHY NOBODY KILLS A COP OR CIA/ DEA OFFICERS IN US..
IT IS A DISGRACE THAT A MUSLIM SHAHRUKH POINTED A GUN AT A COP FROM SIX INCHES RANGE.. AND HE IS STILL ALIVE.. IN ANY OTHER NATION, HE WOULD HAVE BEEN SHOT DEAD ON THE SPOT.. NO JUDGES – NO JURY..
TRAITOR JUDGES IN FOREIGN PAYROLL CAUSED ETHNIC CLEANSING OF KASHMIRI PANDITS...
TRAITOR JUDGES CREATED THE NAXAL RED CORRIDOR...
THEY NEVER EMPHATISED WITH SLAIN JAWANS AND THEIR FAMILIES ..
ILLEGAL COLLEGIUM JUDICIARY HAS NO POWERS TO INTERFERE WITH BHARATMATAs INTERNAL/ EXTERNAL SECURITY....
OUR JUDICIARY IS PACKED WITH TRAITOR JUDGES IN FOREIGN PAYROLL.
WE DONT NEED THE "VISHWAAS" OF TRAITOR MUSLIMS IN PAKISTANI ISI PAYROLL.
https://ajitvadakayil.blogspot.com/2020/01/we-people-are-done-with-illegal.html
BHARATMATA IS BEING BLED BY TRAITOR JUDGES , BENAMI MEDIA JOURNALISTS AND PAKISTANI ISI FUNDED NGOs.
WE THE PEOPLE WATCH IN UTTER FRUSTRATION HOW ILLEGAL COLLEGIUM JUDICIARY IS TREATING “WE THE PEOPLE” AND THE CONSTITUTION IN CONTEMPT..
Capt ajit vadakayil
..
- THIS IS ONE OF THE MOST IMPORTANT COMMENTS EVER MADE ON THIS PLANET.
############################################################
SOMEBODY CALLED ME UP AND ASKED ME
CAPTAIN
WHY IS US DEMOCRAT PRESIDENTIAL CANDIDATE BERNIE SANDERS SUPPORTING MUSLIMS AND RUNNING DOWN HINDUS IN THIS DELHI RIOTS..
WELL
TO UNDERSTAND THIS , LEARN THE FOLLOWING SHOCKING TRUTHS
BERNIE SANDERS IS A COMMIE JEW..
COMMIE JEWS HAVE TAKEN LEADERSHIP OF MUSLIMS ALL OVER THE PLANET SINCE THE PAST SEVERAL CENTURIES ..
JEW ROTHSCHILD CREATED "SEAMLESS BOUNDARIES" IN EU.. WITH COMMON CURRENCY. AFTER THAT HUNDREDS OF THOUSANDS OF SYRIAN MUSLIMS HAVE BEEN ALLOWED TO FLOOD INTO EU AND SCANDINAVIA USING A DROWNED SYRIAN BOY AS A TRIGGER..
WHY?
THE REASON IS ALMOST ALL EU AND SCANDINAVIAN NATIONS ARE RULED BY CRYPTO JEWS..
THE IDEA IS TO SCREW CHRISTIANS / HINDUS USING MUSLIMS ( NAIVE RIGHT TO LEFT WRITING PARTY ) WHOSE TOP LEADERS WILL BE JEWS..
HAVE YOU SEEN A SINGLE TOP MUSLIMS LEADER IN INDIA OR ABROAD WITH ZEBIBA PRAYER MARK IN THE RIGHT PLACE, IF THEY HAVE IT AT ALL?.
http://ajitvadakayil.blogspot.com/2011/07/cracked-heels-and-prayer-marks-capt.html
THE JEWISH DEEP STATE IN ISTANBUL CREATED THE SUNNI/ SHIA DIVIDE..
http://ajitvadakayil.blogspot.com/2019/09/istanbul-deep-seat-of-jewish-deep-state.html
ALL MADRASSAS ON THIS PLANET HAVE BEEN CREATED AND FUNDED BY JEW ROTHSCHILD.. WAHABBI/ SALAFI FUNDS ARE JEWISH..
ISLAMIC BANKING IS JEWISH -- PAKISTANI BANK BCCI WAS A JEWISH BANK. ( I WILL WRITE A FULL POST ON THIS BANK LATER )..
JEW ROTHSCHILD CREATED THE JEWISH PATHAN CLAN ( PASHTUNS )..AND INDUCTED THEM INTO INDIA..
PAKISTANI IMRAN KHAN IS A JEW .. MALALA YOUSAFZAI IS A JEWESS. JINNAH WAS A JEW..
CRYPTO JEW AFRIDI CLAN WAS CREATED TO CONTROL THE BOLAN/ KHYBER PASSES..
EX-PRESIDENT ZAKIR HUSSAIN WHO HAS BEEN SEEN PRAYING IN THE SYNAGOGUE OF HAN MARKET DELHI IS A AFRIDI JEW. HIS GRANDSON IS SALMAN KHURSHID..
ALMOST ALL MAJOR INDIAN NATIONAL CONGRESS MUSLIMS LEADERS WERE JEWS.. MALANA ABDUL KALAM AZAD WAS A QURESHI JEW..
KHAN ABDUL GAFFAR KHAN WAS A JEW.. WE GAVE HIM BHARAT RATNA..
ALMOST ALL MAJOR MUSLIM KINGDOMS IN 1947 WERE RULED BY CRYPTO JEWS.. TIPU SULTAN WAS A JEW. NIZAM OF HYDERABAD IS A JEW..
ISIS WAS CREATED/ ARMED/ FUNDED BY JEWS.. HUNDREDS OF HARDCORE ISLAMIC ISIS SUICIDE BOMBERS NEVER KILLED A SINGLE JEW-- WHY?
THE LAST 70% OF OTTOMAN EMPIRE SULTANS WERE JEWS.. THE MOTHER OF SULTAN MEHMED II WHO FINISHED OFF THE CHRISTIAN ROMAN EMPIRE AT CONSTANTINOPLE WAS A JEWESS.
ALL MOGHUL EMPERORS AFTER HUMAYUN WERE JEWS.. HUMAYUN'S WIFE WAS A JEWESSS.
ALL OIL RICH MUSLIM KINGDOMS OF MIDDLE EAST ARE RULED BY JEW KINGS..
ROTHSCHILD USED JEW LAWRENCE OF ARABIA FOR THIS.. LAWRENCE OF ARABIA WAS MARRIED TO THE DAUGHTER OF FRENCH JEW MICHAEL HARRY NEDOU.. LATER SHEIKH ABDULLAH MARRIED THIS WOMAN AKBAR JEHAN..
WHEN OIL GETS OVER THESE JEW KINGS WILL HAND OVER POWER TO THE ARAB PEOPLE SAYING "DEMOCARASSYY WERY GOOODD" AND RUN AWAY TO THE WEST WHERE THEY HAVE SALTED AWAY THEIR ILL GOTTEN WEALTH ..
POET IQBAL WAS A JEW
JAUHAR ALI BROTHERS WERE JEWS..
JEWS CREATED THE AMU AND JAMIA UNIVERSITIES..
JEW ROTHSCHILD BUILT ALL THE MOSQUES IN KANPUR -- ATTACHED TO HIS TANNERIES.. THESE WERE THE FIRST MOSQUES TO BE FITTED WITH LOUDSPEAKERS, WHICH WERE BASICALLY "RISE AND SHINE REVEILLE CALL" TO START WORKING..
ROTHSCHILD ELIMINATED OTTOMAN SULTANS AND USED JEW MUSTAFA KEMAL ATATURK TO RULE .
ALL YOUNG TURKS WERE JEWS.. IMAGINE THE STUPID INDIAN MEDIA WERE CALLING CHANDRASHEKHAR/ RAHUL GANDHI/ PILOT/ SCINDIA AS YOUNG TURKS..
YOUNG TURK JEWS CONDUCTED THE ARMENIAN CHRISTIAN GENOCIDE AND BLAMED IT ON MUSLIMS..
http://ajitvadakayil.blogspot.com/2015/04/lawrence-of-arabia-part-two-capt-ajit.html
CONTINUED TO 2--
BELOW: SPOT THE CUNT CONTEST
SPOT THE "PIECE OF SHIT" CONTEST
THIS POST IS NOW CONTINUED TO PART 14 BELOW--
https://ajitvadakayil.blogspot.com/2020/02/what-artificial-intelligence-cannot-do_27.html
CAPT AJIT VADAKAYIL
..
SOMEBODY CALLED ME UP AND ASKED ME..
CAPTAIN—
WHO IS MUHAMMAD IBN MUSA AL-KHWARIZMI WHOM MODERN HISTORIANS ARE CALLING THE “FATHER OF COMPUTER SCIENCE” AND THE “FATHER OF ALGORITHMS”??.
LISTEN –
ARAB MUHAMMAD IBN MUSA AL-KHWARIZMI WAS A BRAIN DEAD FELLOW WHOSE ENTIRE WORK WAS SOLD TO HIM TRANSLATED INTO ARABIC BY THE CALCIUT KING FOR GOLD.
THE CALICUT KING MADE HIS MONEY BY NOT ONLY SELLING SPICES –BUT KNOWLEDGE TOO.
HE MAMANKAM FEST HELF AT TIRUNAVAYA KERALA BY THE CALICUT KING EVERY 12 YEARS WAS AN OCCASION WHERE KNOWLEDGE WAS SOLD FOR GOLD.
http://ajitvadakayil.blogspot.com/2019/10/perumal-title-of-calicut-thiyya-kings.html
EVERY ANCIENT GREEK SCHOLAR ( PYTHAGORAS/ PLATO/ SOCRATES ETC ) EXCEPT ARISTOTLE STUDIED AT KODUNGALLUR UNIVERSITY.. THE KERALA SCHOOL OF MATH WAS PART OF IT.
OUR ANCIENT BOOKS ON KNOWLEDGE DID NOT HAVE THE AUTHORs NAME AFFIXED ON THE COVER AS WE CONSIDERED BOOKS AS THE WORK OF SOULS , WHO WOULD BE BORN IN ANOTHER WOMANs WOMB AFTER DEATH.
THE GREEKS TOOK ADVANTAGE OF THIS , STOLE KNOWLEDGE FROM KERALA / INDIA AND PATENTED IT IN THEIR OWN NAMES, WITH HALF BAKED UNDERSTANDING .
WHEN THE KING OF CALICUT CAME TO KNOW THIS, HE BLACKBALLED GREEKS FROM KODUNGALLUR UNIVERSITY .. AND SUDDENLY ANCIENT GREEK KNOWLEDGE DRIED UP LIKE WATER IN THE HOT DESERT SANDS.
LATER THE CALICUT KING SOLD TRANSLATED INTO ARABIC KNOWLEDGE TO BRAIN DEAD ARABS LIKE MUHAMMAD IBN MUSA AL-KHWARIZMI FOR GOLD..
THESE ARAB MIDDLE MEN SOLD KNOWLEDGE ( LIKE MIDDLEMEN FOR SPICES) TO WHITE MEN FOR A PREMIUM.
FIBONACCI TOOK HIS ARABIC WORKS TO ITALY FROM BEJAYA , ALGERIA.
http://ajitvadakayil.blogspot.com/2010/12/perfect-six-pack-capt-ajit-vadakayil.html
EVERY VESTIGE OF ARAB KNOWLEDGE IN THE MIDDLE AGES WAS SOLD IN TRANSLATED ARABIC BY KODUNGALLUR UNIVERSITY FOR GOLD..
FROM 800 AD TO 1450 AD KODUNGALLUR UNIVERSITY OWNED BY THE CALICUT KING EARNED HUGE AMOUNT OF GOLD FOR SELLING READY MADE TRANSLATED KNOWLEDGE ..
THIS IS TIPU SULTANS GOLD WHO STOLE IT FROM NORTH KERALA TEMPLE VAULTS.. ROTHSCHILD BECAME THE RICHEST MAN ON THIS PLANET BY STEALING TIPU SUTANs GOLD IN 1799 AD.
http://ajitvadakayil.blogspot.com/2011/10/tipu-sultan-unmasked-capt-ajit.html
WHEN TIPU SULTAN WAS BLASTING TEMPLE VAULTS, LESS THAN 1% OF THE GOLD WAS SECRETLY TRANSFERRED TO SOUTH KERALA ( TRADITIONAL ENEMIES ) OF THE CALICUT KING. LIKE HOW SADDAM HUSSAIN FLEW HIS FIGHTER JETS TO ENEMY IRAN .
THIS IS THE GOLD WHICH WAS UNEARTHED FROM PADMANABHASWAMY TEMPLE..
http://ajitvadakayil.blogspot.com/2013/01/mansa-musa-king-of-mali-and-sri.html
ALGORITHMS ARE SHORTCUTS PEOPLE USE TO TELL COMPUTERS WHAT TO DO. AT ITS MOST BASIC, AN ALGORITHM SIMPLY TELLS A COMPUTER WHAT TO DO NEXT WITH AN “AND,” “OR,” OR “NOT” STATEMENT.
THE ALGORITHM IS BASICALLY A CODE DEVELOPED TO CARRY OUT A SPECIFIC PROCESS. ALGORITHMS ARE SETS OF RULES, INITIALLY SET BY HUMANS, FOR COMPUTER PROGRAMS TO FOLLOW.
A PROGRAMMING ALGORITHM IS A COMPUTER PROCEDURE THAT IS A LOT LIKE A RECIPE (CALLED A PROCEDURE) AND TELLS YOUR COMPUTER PRECISELY WHAT STEPS TO TAKE TO SOLVE A PROBLEM OR REACH A GOAL.
THERE IS NO ARTIFICIAL INTELLIGENCE WITHOUT ALGORITHMS. ALGORITHMS ARE, IN PART, OUR OPINIONS EMBEDDED IN CODE.
ALGORITHMS ARE AS OLD AS DANAVA CIVILIZATION ITSELF – THIEF GREEK EUCLID’S ALGORITHM BEING ONE OF THE FIRST EXAMPLES DATING BACK SOME 2300 YEARS
EUCLID JUST PATENTED MATH HE LEARNT IN THE KERALA SCHOOL OF MATH IN HIS OWN NAME.. EUCLID IS A THIEF LIKE PYTHAGORAS WHO LEARNT IN THE KERALA SCHOOL OF MATH.
http://ajitvadakayil.blogspot.com/2011/01/isaac-newton-calculus-thief-capt-ajit.html
ALGEBRA DERIVED FROM BRAIN DEAD AL-JABR, ONE OF THE TWO OPERATIONS HE USED TO SOLVE QUADRATIC EQUATIONS.
ALGORISM AND ALGORITHM STEM FROM ALGORITMI, THE LATIN FORM OF HIS NAME.
CONTINUED TO 2--