Artificial intelligence

Early researchers developed algorithms that imitated step-by-step reasoning that humans use when they solve puzzles or make logical deductions.[46] By the late 1980s and 1990s, AI research had developed methods for dealing with uncertain or incomplete information, employing concepts from probability and economics.[47]

Many of these algorithms proved to be insufficient for solving large reasoning problems because they experienced a "combinatorial explosion": they became exponentially slower as the problems grew larger.[48] Even humans rarely use the step-by-step deduction that early AI research could model. They solve most of their problems using fast, intuitive judgments.[49]

An ontology represents knowledge as a set of concepts within a domain and the relationships between those concepts.

Knowledge representation and knowledge engineering[50] allow AI programs to answer questions intelligently and make deductions about real-world facts.

A representation of "what exists" is an ontology: the set of objects, relations, concepts, and properties formally described so that software agents can interpret them.[51] The most general ontologies are called upper ontologies, which attempt to provide a foundation for all other knowledge and act as mediators between domain ontologies that cover specific knowledge about a particular knowledge domain (field of interest or area of concern). A truly intelligent program would also need access to commonsense knowledge; the set of facts that an average person knows. The semantics of an ontology is typically represented in description logic, such as the Web Ontology Language.[52]

AI research has developed tools to represent specific domains, such as objects, properties, categories and relations between objects;[52] situations, events, states and time;[53] causes and effects;[54] knowledge about knowledge (what we know about what other people know);.[55] default reasoning (things that humans assume are true until they are told differently and will remain true even when other facts are changing); [56] as well as other domains. Among the most difficult problems in AI are: the breadth of commonsense knowledge (the number of atomic facts that the average person knows is enormous);[57] and the sub-symbolic form of most commonsense knowledge (much of what people know is not represented as "facts" or "statements" that they could express verbally).[49]

Formal knowledge representations are used in content-based indexing and retrieval,[58] scene interpretation,[59] clinical decision support,[60] knowledge discovery (mining "interesting" and actionable inferences from large databases),[61] and other areas.[62]

An intelligent agent that can plan makes a representation of the state of the world, makes predictions about how their actions will change it and make choices that maximize the utility (or "value") of the available choices.[63] In classical planning problems, the agent can assume that it is the only system acting in the world, allowing the agent to be certain of the consequences of its actions.[64] However, if the agent is not the only actor, then it requires that the agent reason under uncertainty, and continuously re-assess its environment and adapt.[65] Multi-agent planning uses the cooperation and competition of many agents to achieve a given goal. Emergent behavior such as this is used by evolutionary algorithms and swarm intelligence.[66]

Machine learning (ML), a fundamental concept of AI research since the field's inception,[lower-alpha 12] is the study of computer algorithms that improve automatically through experience.[lower-alpha 13]

Unsupervised learning finds patterns in a stream of input. Supervised learning requires a human to label the input data first, and comes in two main varieties: classification and numerical regression. Classification is used to determine what category something belongs in – the program sees a number of examples of things from several categories and will learn to classify new inputs. Regression is the attempt to produce a function that describes the relationship between inputs and outputs and predicts how the outputs should change as the inputs change. Both classifiers and regression learners can be viewed as "function approximators" trying to learn an unknown (possibly implicit) function; for example, a spam classifier can be viewed as learning a function that maps from the text of an email to one of two categories, "spam" or "not spam".[70] In reinforcement learning the agent is rewarded for good responses and punished for bad ones. The agent classifies its responses to form a strategy for operating in its problem space.[71] Transfer learning is when the knowledge gained from one problem is applied to a new problem.[72]

Computational learning theory can assess learners by computational complexity, by sample complexity (how much data is required), or by other notions of optimization.[73]

A parse tree represents the syntactic structure of a sentence according to some formal grammar.

Natural language processing (NLP)[74] allows machines to read and understand human language. A sufficiently powerful natural language processing system would enable natural-language user interfaces and the acquisition of knowledge directly from human-written sources, such as newswire texts. Some straightforward applications of NLP include information retrieval, question answering and machine translation.[75]

Symbolic AI used formal syntax to translate the deep structure of sentences into logic. This failed to produce useful applications, due to the intractability of logic[48] and the breadth of commonsense knowledge.[57] Modern statistical techniques include co-occurrence frequencies (how often one word appears near another), "Keyword spotting" (searching for a particular word to retrieve information), transformer-based deep learning (which finds patterns in text), and others.[76] They have achieved acceptable accuracy at the page or paragraph level, and, by 2019, could generate coherent text.[77]

Feature detection (pictured: edge detection) helps AI compose informative abstract structures out of raw data.

Machine perception[78] is the ability to use input from sensors (such as cameras, microphones, wireless signals, and active lidar, sonar, radar, and tactile sensors) to deduce aspects of the world. Applications include speech recognition,[79] facial recognition, and object recognition.[80] Computer vision is the ability to analyze visual input.[81]

AI is heavily used in robotics.[82] Localization is how a robot knows its location and maps its environment. When given a small, static, and visible environment, this is easy; however, dynamic environments, such as (in endoscopy) the interior of a patient's breathing body, pose a greater challenge.[83]

Motion planning is the process of breaking down a movement task into "primitives" such as individual joint movements. Such movement often involves compliant motion, a process where movement requires maintaining physical contact with an object. Robots can learn from experience how to move efficiently despite the presence of friction and gear slippage.[84]

Kismet, a robot with rudimentary social skills[85]

Affective computing is an interdisciplinary umbrella that comprises systems that recognize, interpret, process or simulate human feeling, emotion and mood.[86] For example, some virtual assistants are programmed to speak conversationally or even to banter humorously; it makes them appear more sensitive to the emotional dynamics of human interaction, or to otherwise facilitate human–computer interaction. However, this tends to give naïve users an unrealistic conception of how intelligent existing computer agents actually are.[87] Moderate successes related to affective computing include textual sentiment analysis and, more recently, multimodal sentiment analysis), wherein AI classifies the affects displayed by a videotaped subject.[88]

A machine with general intelligence can solve a wide variety of problems with breadth and versatility similar to human intelligence. There are several competing ideas about how to develop artificial general intelligence. Hans Moravec and Marvin Minsky argue that work in different individual domains can be incorporated into an advanced multi-agent system or cognitive architecture with general intelligence.[89] Pedro Domingos hopes that there is a conceptually straightforward, but mathematically difficult, "master algorithm" that could lead to AGI.[90] Others believe that anthropomorphic features like an artificial brain[91] or simulated child development[lower-alpha 14] will someday reach a critical point where general intelligence emerges.

AI can solve many problems by intelligently searching through many possible solutions.[92] Reasoning can be reduced to performing a search. For example, logical proof can be viewed as searching for a path that leads from premises to conclusions, where each step is the application of an inference rule.[93] Planning algorithms search through trees of goals and subgoals, attempting to find a path to a target goal, a process called means-ends analysis.[94] Robotics algorithms for moving limbs and grasping objects use local searches in configuration space.[95]

Simple exhaustive searches[96] 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 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 eliminate some choices 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.[97] Heuristics limit the search for solutions into a smaller sample size.[98]

A particle swarm seeking the global minimum

A very different kind of search came to prominence in the 1990s, based on the mathematical theory of optimization. For many problems, it is possible to begin the search with some form of a guess and then refine the guess incrementally until no more refinements can be made. These algorithms can be visualized as blind hill climbing: we begin the search at a random point on the landscape, and then, by jumps or steps, we keep moving our guess uphill, until we reach the top. Other related optimization algorithms include random optimization, beam search and metaheuristics like simulated annealing.[99] Evolutionary computation uses a form of optimization search. For example, they may begin with a population of organisms (the guesses) and then allow them to mutate and recombine, selecting only the fittest to survive each generation (refining the guesses). Classic evolutionary algorithms include genetic algorithms, gene expression programming, and genetic programming.[100] Alternatively, distributed search processes can coordinate via swarm intelligence algorithms. Two popular swarm algorithms used in search are particle swarm optimization (inspired by bird flocking) and ant colony optimization (inspired by ant trails).[101]

Logic[102] is used for knowledge representation and problem-solving, but it can be applied to other problems as well. For example, the satplan algorithm uses logic for planning[103] and inductive logic programming is a method for learning.[104]

Several different forms of logic are used in AI research. Propositional logic[105] involves truth functions such as "or" and "not". First-order logic[106] adds quantifiers and predicates and can express facts about objects, their properties, and their relations with each other. Fuzzy logic assigns a "degree of truth" (between 0 and 1) to vague statements such as "Alice is old" (or rich, or tall, or hungry), that are too linguistically imprecise to be completely true or false.[107] Default logics, non-monotonic logics and circumscription are forms of logic designed to help with default reasoning and the qualification problem.[56] Several extensions of logic have been designed to handle specific domains of knowledge, such as description logics;[52] situation calculus, event calculus and fluent calculus (for representing events and time);[53] causal calculus;[54] belief calculus (belief revision); and modal logics.[55] Logics to model contradictory or inconsistent statements arising in multi-agent systems have also been designed, such as paraconsistent logics.[108]

Expectation-maximization clustering of Old Faithful eruption data starts from a random guess but then successfully converges on an accurate clustering of the two physically distinct modes of eruption.

Many problems in AI (including in reasoning, planning, learning, perception, and robotics) require the agent to operate with incomplete or uncertain information. AI researchers have devised a number of tools to solve these problems using methods from probability theory and economics.[109] Bayesian networks[110] are a very general tool that can be used for various problems, including reasoning (using the Bayesian inference algorithm),[lower-alpha 15][112] learning (using the expectation-maximization algorithm),[lower-alpha 16][114] planning (using decision networks)[115] and perception (using dynamic Bayesian networks).[116] Probabilistic algorithms can also be used for filtering, prediction, smoothing and finding explanations for streams of data, helping perception systems to analyze processes that occur over time (e.g., hidden Markov models or Kalman filters).[116]

A key concept from the science of economics is "utility", a measure of how valuable something is to an intelligent agent. Precise mathematical tools have been developed that analyze how an agent can make choices and plan, using decision theory, decision analysis,[117] and information value theory.[118] These tools include models such as Markov decision processes,[119] dynamic decision networks,[116] game theory and mechanism design.[120]

The simplest AI applications can be divided into two types: classifiers ("if shiny then diamond") and controllers ("if diamond then pick up"). Controllers do, however, also classify conditions before inferring actions, and therefore classification forms a central part of many AI systems. Classifiers are functions that use pattern matching to determine the closest match. They can be tuned according to examples, making them very attractive for use in AI. These examples are known as observations or patterns. In supervised learning, each pattern belongs to a certain predefined class. A class is a decision that has to be made. All the observations combined with their class labels are known as a data set. When a new observation is received, that observation is classified based on previous experience.[121]

A classifier can be trained in various ways; there are many statistical and machine learning approaches. The decision tree is the simplest and most widely used symbolic machine learning algorithm.[122] K-nearest neighbor algorithm was the most widely used analogical AI until the mid-1990s.[123] Kernel methods such as the support vector machine (SVM) displaced k-nearest neighbor in the 1990s.[124] The naive Bayes classifier is reportedly the "most widely used learner"[125] at Google, due in part to its scalability.[126] Neural networks are also used for classification.[127]

Classifier performance depends greatly on the characteristics of the data to be classified, such as the dataset size, distribution of samples across classes, dimensionality, and the level of noise. Model-based classifiers perform well if the assumed model is an extremely good fit for the actual data. Otherwise, if no matching model is available, and if accuracy (rather than speed or scalability) is the sole concern, conventional wisdom is that discriminative classifiers (especially SVM) tend to be more accurate than model-based classifiers such as "naive Bayes" on most practical data sets.[128]

A neural network is an interconnected group of nodes, akin to the vast network of neurons in the human brain.

Neural networks[127] were inspired by the architecture of neurons in the human brain. A simple "neuron" N accepts input from other neurons, each of which, when activated (or "fired"), casts a weighted "vote" for or against whether neuron N should itself activate. Learning requires an algorithm to adjust these weights based on the training data; one simple algorithm (dubbed "fire together, wire together") is to increase the weight between two connected neurons when the activation of one triggers the successful activation of another. Neurons have a continuous spectrum of activation; in addition, neurons can process inputs in a nonlinear way rather than weighing straightforward votes.

Modern neural networks model complex relationships between inputs and outputs and find patterns in data. They can learn continuous functions and even digital logical operations. Neural networks can be viewed as a type of mathematical optimization – they perform gradient descent on a multi-dimensional topology that was created by training the network. The most common training technique is the backpropagation algorithm.[129] Other learning techniques for neural networks are Hebbian learning ("fire together, wire together"), GMDH or competitive learning.[130]

The main categories of networks are acyclic or feedforward neural networks (where the signal passes in only one direction) and recurrent neural networks (which allow feedback and short-term memories of previous input events). Among the most popular feedforward networks are perceptrons, multi-layer perceptrons and radial basis networks.[131]

Representing images on multiple layers of abstraction in deep learning[132]

Deep learning[133] uses several layers of neurons between the network's inputs and outputs. The multiple layers can progressively extract higher-level features from the raw input. For example, in image processing, lower layers may identify edges, while higher layers may identify the concepts relevant to a human such as digits or letters or faces.[134] Deep learning has drastically improved the performance of programs in many important subfields of artificial intelligence, including computer vision, speech recognition, image classification[135] and others.

Deep learning often uses convolutional neural networks for many or all of its layers. In a convolutional layer, each neuron receives input from only a restricted area of the previous layer called the neuron's receptive field. This can substantially reduce the number of weighted connections between neurons,[136] and creates a hierarchy similar to the organization of the animal visual cortex.[137]

In a recurrent neural network (RNN) the signal will propagate through a layer more than once;[138] thus, an RNN is an example of deep learning.[139] RNNs can be trained by gradient descent,[140] however long-term gradients which are back-propagated can "vanish" (that is, they can tend to zero) or "explode" (that is, they can tend to infinity), known as the vanishing gradient problem.[141] The long short term memory (LSTM) technique can prevent this in most cases.[142]

Specialized languages for artificial intelligence have been developed, such as Lisp, Prolog, TensorFlow and many others. Hardware developed for AI includes AI accelerators and neuromorphic computing.

Alan Turing wrote in 1950 "I propose to consider the question 'can machines think'?"[163] He advised changing the question from whether a machine "thinks", to "whether or not it is possible for machinery to show intelligent behaviour".[164] He devised the Turing test, which measures the ability of a machine to simulate human conversation.[165]

Oxford English Dictionary (Oxford University Press) defines artificial intelligence as:[166]

the theory and development of computer systems able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages.

A definition that has also been adopted by Google[167] - major practitionary in the field of AI - and other AI practitioners.

This definition stipulated the ability of systems to synthesize information as the manifestation of intelligence, similar to the way it is defined in biological intelligence. It does not require a procedural-action component, such as in many definitions of intelligent agents.

For many fields such as law, political science, and other social science, it is often argued that for purposes in these fields only instead of trying to arrive at a rigid definition of AI, it is often easier to work with the paradigm of intelligent agents. It should be noted that these are working definitions of broader systems-that-use-artificial-intelligence (sometimes called: AI-based systems), but that these are not rigid definitions of AI itself.

The only thing visible is the behavior of the machine, so it does not matter if the machine is conscious, or has a mind, or whether the intelligence is merely a "simulation" and not "the real thing".[lower-alpha 18]

Russell and Norvig are critical of the Turing test if it is used as the definition of artificial intelligence. "Aeronautical engineering texts," they wrote, "do not define the goal of their field as making 'machines that fly so exactly like pigeons that they can fool other pigeons.'"[169] AI founder John McCarthy agreed, writing that "Artificial intelligence is not, by definition, simulation of human intelligence".[170][lower-alpha 19]

The intelligent agent paradigm[172] defines intelligent behavior in general, without reference to human beings. An intelligent agent is a system that perceives its environment and takes actions that maximize its chances of success. Any system that has goal-directed behavior can be analyzed as an intelligent agent: something as simple as a thermostat, as complex as a human being, as well as large systems such as firms, biomes or nations. The intelligent agent paradigm became widely accepted during the 1990s, and currently serves as the definition of the field.[lower-alpha 1]

The intelligent agent paradigm has other advantages for AI. It provides a reliable and scientific way to test programs; researchers can directly compare or even combine different approaches to isolated problems, by asking which agent is best at maximizing a given "goal function". It also gives them a common language to communicate with other fields – such as mathematical optimization (which is defined in terms of "goals") or economics (which uses the same definition of a "rational agent").[173]

No established unifying theory or paradigm has guided AI research for most of its history.[lower-alpha 20] The unprecedented success of statistical machine learning in the 2010s eclipsed all other approaches (so much so that some sources, especially in the business world, use the term "artificial intelligence" to mean "machine learning with neural networks"). This approach is mostly sub-symbolic, neat, soft and narrow (see below). Critics argue that these questions may have to be revisited by future generations of AI researchers.

Symbolic AI (or "GOFAI")[175] simulated the high-level conscious reasoning that people use when they solve puzzles, express legal reasoning and do mathematics. They were highly successful at "intelligent" tasks such as algebra or IQ tests. In the 1960s, Newell and Simon proposed the physical symbol systems hypothesis: "A physical symbol system has the necessary and sufficient means of general intelligent action."[176]

However, the symbolic approach failed on many tasks that humans solve easily, such as learning, recognizing an object or commonsense reasoning. Moravec's paradox is the discovery that high-level "intelligent" tasks were easy for AI, but low level "instinctive" tasks were extremely difficult.[177] Philosopher Hubert Dreyfus had argued since the 1960s that human expertise depends on unconscious instinct rather than conscious symbol manipulation, and on having a "feel" for the situation, rather than explicit symbolic knowledge.[178] Although his arguments had been ridiculed and ignored when they were first presented, eventually, AI research came to agree.[lower-alpha 21][49]

The issue is not resolved: sub-symbolic reasoning can make many of the same inscrutable mistakes that human intuition does, such as algorithmic bias. Critics such as Noam Chomsky argue continuing research into symbolic AI will still be necessary to attain general intelligence,[180][181] in part because sub-symbolic AI is a move away from explainable AI: it can be difficult or impossible to understand why a modern statistical AI program made a particular decision. The emerging field of neurosymbolic artificial intelligence attempts to bridge the two approaches.

"Neats" hope that intelligent behavior is described using simple, elegant principles (such as logic, optimization, or neural networks). "Scruffies" expect that it necessarily requires solving a large number of unrelated problems (especially in areas like common sense reasoning). This issue was actively discussed in the 70s and 80s,[182] but in the 1990s mathematical methods and solid scientific standards became the norm, a transition that Russell and Norvig termed "the victory of the neats".[183]

Finding a provably correct or optimal solution is intractable for many important problems.[48] Soft computing is a set of techniques, including genetic algorithms, fuzzy logic and neural networks, that are tolerant of imprecision, uncertainty, partial truth and approximation. Soft computing was introduced in the late 80s and most successful AI programs in the 21st century are examples of soft computing with neural networks.

AI researchers are divided as to whether to pursue the goals of artificial general intelligence and superintelligence (general AI) directly or to solve as many specific problems as possible (narrow AI) in hopes these solutions will lead indirectly to the field's long-term goals.[184][185] General intelligence is difficult to define and difficult to measure, and modern AI has had more verifiable successes by focusing on specific problems with specific solutions. The experimental sub-field of artificial general intelligence studies this area exclusively.

The philosophy of mind does not know whether a machine can have a mind, consciousness and mental states, in the same sense that human beings do. This issue considers the internal experiences of the machine, rather than its external behavior. Mainstream AI research considers this issue irrelevant because it does not affect the goals of the field. Stuart Russell and Peter Norvig observe that most AI researchers "don't care about the [philosophy of AI] – as long as the program works, they don't care whether you call it a simulation of intelligence or real intelligence."[186] However, the question has become central to the philosophy of mind. It is also typically the central question at issue in artificial intelligence in fiction.

David Chalmers identified two problems in understanding the mind, which he named the "hard" and "easy" problems of consciousness.[187] The easy problem is understanding how the brain processes signals, makes plans and controls behavior. The hard problem is explaining how this feels or why it should feel like anything at all. Human information processing is easy to explain, however, human subjective experience is difficult to explain. For example, it is easy to imagine a color-blind person who has learned to identify which objects in their field of view are red, but it is not clear what would be required for the person to know what red looks like.[188]

Computationalism is the position in the philosophy of mind that the human mind is an information processing system and that thinking is a form of computing. Computationalism argues that the relationship between mind and body is similar or identical to the relationship between software and hardware and thus may be a solution to the mind-body problem. This philosophical position was inspired by the work of AI researchers and cognitive scientists in the 1960s and was originally proposed by philosophers Jerry Fodor and Hilary Putnam.[189]

Philosopher John Searle characterized this position as "strong AI": "The appropriately programmed computer with the right inputs and outputs would thereby have a mind in exactly the same sense human beings have minds."[lower-alpha 22] Searle counters this assertion with his Chinese room argument, which attempts to show that, even if a machine perfectly simulates human behavior, there is still no reason to suppose it also has a mind.[192]

If a machine has a mind and subjective experience, then it may also have sentience (the ability to feel), and if so, then it could also suffer, and thus it would be entitled to certain rights.[193] Any hypothetical robot rights would lie on a spectrum with animal rights and human rights.[194] This issue has been considered in fiction for centuries,[195] and is now being considered by, for example, California's Institute for the Future; however, critics argue that the discussion is premature.[196]

A superintelligence, hyperintelligence, or superhuman intelligence, is a hypothetical agent that would possess intelligence far surpassing that of the brightest and most gifted human mind. Superintelligence may also refer to the form or degree of intelligence possessed by such an agent.[185]

If research into artificial general intelligence produced sufficiently intelligent software, it might be able to reprogram and improve itself. The improved software would be even better at improving itself, leading to recursive self-improvement.[197] Its intelligence would increase exponentially in an intelligence explosion and could dramatically surpass humans. Science fiction writer Vernor Vinge named this scenario the "singularity".[198] Because it is difficult or impossible to know the limits of intelligence or the capabilities of superintelligent machines, the technological singularity is an occurrence beyond which events are unpredictable or even unfathomable.[199]

Robot designer Hans Moravec, cyberneticist Kevin Warwick, and inventor Ray Kurzweil have predicted that humans and machines will merge in the future into cyborgs that are more capable and powerful than either. This idea, called transhumanism, has roots in Aldous Huxley and Robert Ettinger.[200]

Edward Fredkin argues that "artificial intelligence is the next stage in evolution", an idea first proposed by Samuel Butler's "Darwin among the Machines" as far back as 1863, and expanded upon by George Dyson in his book of the same name in 1998.[201]

In the past, technology has tended to increase rather than reduce total employment, but economists acknowledge that "we're in uncharted territory" with AI.[202] A survey of economists showed disagreement about whether the increasing use of robots and AI will cause a substantial increase in long-term unemployment, but they generally agree that it could be a net benefit if productivity gains are redistributed.[203] Subjective estimates of the risk vary widely; for example, Michael Osborne and Carl Benedikt Frey estimate 47% of U.S. jobs are at "high risk" of potential automation, while an OECD report classifies only 9% of U.S. jobs as "high risk".[lower-alpha 23][205]

Unlike previous waves of automation, many middle-class jobs may be eliminated by artificial intelligence; The Economist states that "the worry that AI could do to white-collar jobs what steam power did to blue-collar ones during the Industrial Revolution" is "worth taking seriously".[206] Jobs at extreme risk range from paralegals to fast food cooks, while job demand is likely to increase for care-related professions ranging from personal healthcare to the clergy.[207]

AI provides a number of tools that are particularly useful for authoritarian governments: smart spyware, face recognition and voice recognition allow widespread surveillance; such surveillance allows machine learning to classify potential enemies of the state and can prevent them from hiding; recommendation systems can precisely target propaganda and misinformation for maximum effect; deepfakes aid in producing misinformation; advanced AI can make centralized decision making more competitive with liberal and decentralized systems such as markets.[208]

Terrorists, criminals and rogue states may use other forms of weaponized AI such as advanced digital warfare and lethal autonomous weapons. By 2015, over fifty countries were reported to be researching battlefield robots.[209]

Machine-learning AI is also able to design tens of thousands of toxic molecules in a matter of hours.[210]

AI programs can become biased after learning from real-world data. It is not typically introduced by the system designers but is learned by the program, and thus the programmers are often unaware that the bias exists.[211] Bias can be inadvertently introduced by the way training data is selected.[212] It can also emerge from correlations: AI is used to classify individuals into groups and then make predictions assuming that the individual will resemble other members of the group. In some cases, this assumption may be unfair.[213] An example of this is COMPAS, a commercial program widely used by U.S. courts to assess the likelihood of a defendant becoming a recidivist. ProPublica claims that the COMPAS-assigned recidivism risk level of black defendants is far more likely to be overestimated than that of white defendants, despite the fact that the program was not told the races of the defendants.[214] Other examples where algorithmic bias can lead to unfair outcomes are when AI is used for credit rating or hiring.

At its 2022 Conference on Fairness, Accountability, and Transparency (ACM FAccT 2022) the Association for Computing Machinery, in Seoul, South Korea, presented and published findings recommending that until AI and robotics systems are demonstrated to be free of bias mistakes, they are unsafe and the use of self-learning neural networks trained on vast, unregulated sources of flawed internet data should be curtailed.[215]

Superintelligent AI may be able to improve itself to the point that humans could not control it. This could, as physicist Stephen Hawking puts it, "spell the end of the human race".[216] Philosopher Nick Bostrom argues that sufficiently intelligent AI, if it chooses actions based on achieving some goal, will exhibit convergent behavior such as acquiring resources or protecting itself from being shut down. If this AI's goals do not fully reflect humanity's, it might need to harm humanity to acquire more resources or prevent itself from being shut down, ultimately to better achieve its goal. He concludes that AI poses a risk to mankind, however humble or "friendly" its stated goals might be.[217] Political scientist Charles T. Rubin argues that "any sufficiently advanced benevolence may be indistinguishable from malevolence." Humans should not assume machines or robots would treat us favorably because there is no a priori reason to believe that they would share our system of morality.[218]

The opinion of experts and industry insiders is mixed, with sizable fractions both concerned and unconcerned by risk from eventual superhumanly-capable AI.[219] Stephen Hawking, Microsoft founder Bill Gates, history professor Yuval Noah Harari, and SpaceX founder Elon Musk have all expressed serious misgivings about the future of AI.[220] Prominent tech titans including Peter Thiel (Amazon Web Services) and Musk have committed more than $1 billion to nonprofit companies that champion responsible AI development, such as OpenAI and the Future of Life Institute.[221] Mark Zuckerberg (CEO, Facebook) has said that artificial intelligence is helpful in its current form and will continue to assist humans.[222] Other experts argue is that the risks are far enough in the future to not be worth researching, or that humans will be valuable from the perspective of a superintelligent machine.[223] Rodney Brooks, in particular, has said that "malevolent" AI is still centuries away.[lower-alpha 24]

AI's decisions making abilities raises the questions of legal responsibility and copyright status of created works. This issues are being refined in various jurisdictions.[225]

Friendly AI are machines that have been designed from the beginning to minimize risks and to make choices that benefit humans. Eliezer Yudkowsky, who coined the term, argues that developing friendly AI should be a higher research priority: it may require a large investment and it must be completed before AI becomes an existential risk.[226]

Machines with intelligence have the potential to use their intelligence to make ethical decisions. The field of machine ethics provides machines with ethical principles and procedures for resolving ethical dilemmas.[227] Machine ethics is also called machine morality, computational ethics or computational morality,[227] and was founded at an AAAI symposium in 2005.[228]

Other approaches include Wendell Wallach's "artificial moral agents"[229] and Stuart J. Russell's three principles for developing provably beneficial machines.[230]

The regulation of artificial intelligence is the development of public sector policies and laws for promoting and regulating artificial intelligence (AI); it is therefore related to the broader regulation of algorithms.[231] The regulatory and policy landscape for AI is an emerging issue in jurisdictions globally.[232] Between 2016 and 2020, more than 30 countries adopted dedicated strategies for AI.[45] Most EU member states had released national AI strategies, as had Canada, China, India, Japan, Mauritius, the Russian Federation, Saudi Arabia, United Arab Emirates, US and Vietnam. Others were in the process of elaborating their own AI strategy, including Bangladesh, Malaysia and Tunisia.[45] The Global Partnership on Artificial Intelligence was launched in June 2020, stating a need for AI to be developed in accordance with human rights and democratic values, to ensure public confidence and trust in the technology.[45] Henry Kissinger, Eric Schmidt, and Daniel Huttenlocher published a joint statement in November 2021 calling for a government commission to regulate AI.[233]

As technology and research evolve and the world enters the third revolution of warfare following gunpowder and nuclear weapons, the artificial intelligence arms race ensues between the United States, China, and Russia, three countries with the world's top five highest military budgets.[238] Intentions of being a world leader in AI research by 2030[239] have been declared by China's leader Xi Jinping, and President Putin of Russia has stated that "Whoever becomes the leader in this sphere will become the ruler of the world".[240] If Russia were to become the leader in AI research, President Putin has stated Russia's intent to share some of their research with the world so as to not monopolize the field,[240] similar to their current sharing of nuclear technologies, maintaining science diplomacy relations. The United States, China, and Russia, are some examples of countries that have taken their stances toward military artificial intelligence since as early as 2014, having established military programs to develop cyber weapons, control lethal autonomous weapons, and drones that can be used for surveillance.

President Putin announced that artificial intelligence is the future for all mankind [240] and recognizes the power and opportunities that the development and deployment of lethal autonomous weapons AI technology can hold in warfare and homeland security, as well as its threats. President Putin's prediction that future wars will be fought using AI has started to come to fruition to an extent after Russia invaded Ukraine on 24 February 2022.  The Ukrainian military is making use of the Turkish Bayraktar TB2-drones[241] that still require human operation to deploy laser-guided bombs but can take off, land, and cruise autonomously. Ukraine has also been using Switchblade drones supplied by the US and receiving information gathering by the United States's own surveillance operations regarding battlefield intelligence and national security about Russia.[242] Similarly, Russia can use AI to help analyze battlefield data from surveillance footage taken by drones. Reports and images show that Russia's military has deployed KUB- BLA suicide drones [243] into Ukraine, with speculations of intentions to assassinate Ukrainian President Volodymyr Zelenskyy.

As research in the AI realm progresses, there is pushback about the use of AI from the Campaign to Stop Killer Robots and world technology leaders have sent a petition[244] to the United Nations calling for new regulations on the development and use of AI technologies in 2017, including a ban on the use of lethal autonomous weapons due to ethical concerns for innocent civilian populations.

With the ever evolving cyber-attacks and generation of devices, AI can be used for threat detection and more effective response by risk prioritization. With this tool, some challenges are also presented such as privacy, informed consent, and responsible use.[245] According to CISA, the cyberspace is difficult to secure for the following factors: the ability of malicious actors to operate from anywhere in the world, the linkages between cyberspace and physical systems, and the difficulty of reducing vulnerabilities and consequences in complex cyber networks.[246] With the increased technological advances of the world, the risk for wide scale consequential events rises. Paradoxically, the ability to protect information and create a line of communication between the scientific and diplomatic community thrives. The role of cybersecurity in diplomacy has become increasingly relevant, creating the term of cyber diplomacy – which is not uniformly defined and not synonymous with cyber defence.[247] Many nations have developed unique approaches to scientific diplomacy in cyberspace.

Dating back to 2011, when the Czech National Security Authority (NSA) was appointed as the national authority for the cyber agenda. The role of cyber diplomacy strengthened in 2017 when the Czech Ministry of Foreign Affairs (MFA) detected a serious cyber campaign directed against its own computer networks.[248] In 2016, three cyber diplomats were deployed to Washington, D.C., Brussels and Tel Aviv, with the goal of establishing active international cooperation focused on engagement with the EU and NATO. The main agenda for these scientific diplomacy efforts is to bolster research on artificial intelligence and how it can be utilized in cybersecurity research, development, and overall consumer trust.[249] CzechInvest is a key stakeholder in scientific diplomacy and cybersecurity. For example, in September 2018, they organized a mission to Canada in September 2018 with a special focus on artificial intelligence. The main goal of this particular mission was a promotional effort on behalf of Prague, attempting to establish it as a future knowledge hub for the industry for interested Canadian firms.[250]

Cybersecurity is recognized as a governmental task, dividing into three ministries of responsibility: the Federal Ministry of the Interior, the Federal Ministry of Defence, and the Federal Foreign Office.[251] These distinctions promoted the creation of various institutions, such as The German National Office for Information Security, The National Cyberdefence Centre, The German National Cyber Security Council, and The Cyber and Information Domain Service.[249] In 2018, a new strategy for artificial intelligence was established by the German government, with the creation of a German-French virtual research and innovation network,[252] holding opportunity for research expansion into cybersecurity.

The adoption of The Cybersecurity Strategy of the European Union – An Open, Safe and Secure Cyberspace document in 2013 by the European commission[249] pushed forth cybersecurity efforts integrated with scientific diplomacy and artificial intelligence. Efforts are strong, as the EU funds various programs and institutions in the effort to bring science to diplomacy and bring diplomacy to science. Some examples are the cyber security programme Competence Research Innovation (CONCORDIA), which brings together 14 member states,[253] and Cybersecurity for Europe (CSE), which brings together 43 partners involving 20 member states.[254] In addition, The European Network of Cybersecurity Centres and Competence Hub for Innovation and Operations (ECHO) gathers 30 partners with 15 member states[255] and SPARTA gathers 44 partners involving 14 member states.[256] These efforts reflect the overall goals of the EU, to innovate cybersecurity for defense and protection, establish a highly integrated cyberspace among many nations, and further contribute to the security of artificial intelligence.[249]

With the 2022 invasion of Ukraine, there has been a rise in malicious cyber activity against the United States,[257] Ukraine, and Russia. A prominent and rare documented use of artificial intelligence in conflict is on behalf of Ukraine, using facial recognition software to uncover Russian assailants and identify Ukrainians killed in the ongoing war.[258] Though these governmental figures are not primarily focused on scientific and cyber diplomacy, other institutions are commenting on the use of artificial intelligence in cybersecurity with that focus. For example, Georgetown University's Center for Security and Emerging Technology (CSET) has the Cyber-AI Project, with one goal being to attract policymakers' attention to the growing body of academic research, whichexposes the exploitive consequences of AI and machine-learning (ML) algorithms.[259] This vulnerability can be a plausible explanation as to why Russia is not engaging in the use of AI in conflict per, Andrew Lohn, a senior fellow at CSET. In addition to use on the battlefield, AI is being used by the Pentagon to analyze data from the war, analyzing to strengthen cybersecurity and warfare intelligence for the United States.[242][260]

As artificial intelligence grows and the overwhelming amount of news portrayed through cyberspace expands, it is becoming extremely overwhelming for a voter to know what to believe. There are many intelligent codes, referred to as bots, written to portray people on social media with the goal of spreading miss information.[261] The 2016 US election is a victim of such actions. During the Hillary Clinton and Donald Trump campaign, artificial intelligent bots from Russia were spreading misinformation about the candidates in order to help the Trump campaign.[262] Analysts concluded that approximately 19% of Twitter tweets centered around the 2016 election were detected to come from bots.[262] YouTube in recent years has been used to spread political information as well. Although there is no proof that the platform attempts to manipulate its viewers opinions, Youtubes AI algorithm recommends videos of similar variety.[263] If a person begins to research right wing political podcasts, then YouTube's algorithm will recommend more right wing videos.[264] The uprising in a program called Deepfake, a software used to replicate someone's face and words, has also shown its potential threat. In 2018 a Deepfake video of Barack Obama was released saying words he claims to have never said.[265] While in a national election a Deepfake will quickly be debunked, the software has the capability to heavily sway a smaller local election. This tool holds a lot of potential for spreading misinformation and is monitored with great attention.[266] Although it may be seen as a tool used for harm, AI can help enhance election campaigns as well. AI bots can be programed to target articles with known misinformation. The bots can then indicate what is being misinformed to help shine light on the truth. AI can also be used to inform a person where each parts stands on a certain topic such as healthcare or climate change.[267] The political leaders of a nation have heavy sway on international affairs. Thus, a political leader with a lack of interest for international collaborative scientific advancement can have a negative impact in the scientific diplomacy of that nation[268]

The use of artificial intelligence (AI) has subtly grown to become part of everyday life. It is used every day in facial recognition software. It is the first measure of security for many companies in the form of a biometric authentication. This means of authentication allows even the most official organizations such as the United States Internal Revenue Service to verify a person's identity [269] via a database generated from machine learning. As of the year 2022, the United States IRS requires those who do not undergo a live interview with an agent to complete a biometric verification of their identity via ID.me's facial recognition tool.[269]

In Japan and South Korea, artificial intelligence software is used in the instruction of English language via the company Riiid.[270] Riiid is a Korean education company working alongside Japan to give students the means to learn and use their English communication skills via engaging with artificial intelligence in a live chat.[270] Riid is not the only company to do this. American company Duolingo is well known for their automated teaching of 41 languages. Babbel, a German language learning program, also uses artificial intelligence in its teaching automation, allowing for European students to learn vital communication skills needed in social, economic, and diplomatic settings. Artificial intelligence will also automate the routine tasks that teachers need to do such as grading, taking attendance, and handling routine student inquiries.[271] This enables the teacher to carry on with the complexities of teaching that an automated machine cannot handle. These include creating exams, explaining complex material in a way that will benefit students individually and handling unique questions from students.

Unlike the human brain, which possess generalized intelligence, the specialized intelligence of AI can serve as a means of support to physicians internationally. The medical field has a diverse and profound amount of data in which AI can employ to generate a predictive diagnosis. Researchers at an Oxford hospital have developed artificial intelligence that can diagnose heart scans for heart disease and cancer.[272] This artificial intelligence can pick up diminutive details in the scans that doctors may miss. As such, artificial intelligence in medicine will better the industry, giving doctors the means to precisely diagnose their patients using the tools available. The artificial intelligence algorithms will also be used to further improve diagnosis over time, via an application of machine learning called precision medicine.[273] Furthermore, the narrow application of artificial intelligence can use "deep learning" in order to improve medical image analysis. In radiology imaging, AI uses deep learning algorithms to identify potentially cancerous lesions which is an important process assisting in early diagnosis.[274]

Data analysis is a fundamental property of artificial intelligence that enables it to be used in every facet of life from search results to the way people buy product. According to NewVantage Partners,[275] over 90% of top businesses have ongoing investments in artificial intelligence. According to IBM, one of the world's leaders in technology, 45% of respondents from companies with over 1,000 employees have adopted AI.[276] Recent data shows that the business market [277] for artificial intelligence during the year 2020 was valued at $51.08 billion. The business market for artificial intelligence is projected to be over $640.3 billion by the year 2028.[277] To prevent harm, AI-deploying organizations need to play a central role in creating and deploying trustworthy AI in line with the principles of trustworthy AI,[278] and take accountability to mitigate the risks.[279]

With the exponential surge of artificial technology and communication, the distribution of one's ideals and values has been evident in daily life. Digital information is spread via communication apps such as Whatsapp, Facebook/Meta, Snapchat, Instagram and Twitter. However, it is known that these sites relay specific information corresponding to data analysis. If a right-winged individual were to do a google search, Google's algorithms would target that individual and relay data pertinent to that target audience. US President Bill Clinton noted in 2000:"In the new century, liberty will spread by cell phone and cable modem. [...] We know how much the Internet has changed America, and we are already an open society.[280] However, when the private sector uses artificial intelligence to gather data, a shift in power from the state to the private sector may be seen. This shift in power, specifically in large technological corporations, could profoundly change how diplomacy functions in society. The rise in digital technology and usage of artificial technology enabled the private sector to gather immense data on the public, which is then further categorized by race, location, age, gender, etc.[281] The New York Times calculates that "the ten largest tech firms, which have become gatekeepers in commerce, finance, entertainment and communications, now have a combined market capitalization of more than $10 trillion. In gross domestic product terms, that would rank them as the world's third-largest economy."[282] Beyond the general lobbying of congressmen/congresswomen, companies such as Facebook/Meta or Google use collected data in order to reach their intended audiences with targeted information.[282]

Multiple nations around the globe employ artificial intelligence to assist with their foreign policy decisions. The Chinese Department of External Security Affairs – under the Ministry of Foreign Affairs – uses AI to review almost all its foreign investment projects for risk mitigation.[283] The government of China plans to utilize artificial intelligence in its $900 billion global infrastructure development plan, called the "Belt and Road Initiative" for political, economic, and environmental risk alleviation.[284]

Over 200 applications of artificial intelligence are being used by over 46 United Nations agencies, in sectors ranging from health care dealing with issues such as combating COVID-19 to smart agriculture, to assist the UN in political and diplomatic relations.[285] One example is the use of AI by the UN Global Pulse program to model the effect of the spread of COVID-19 on internally displaced people (IDP) and refugee settlements to assist them in creating an appropriate global health policy.[286][287]

Novel AI tools such as remote sensing can also be employed by diplomats for collecting and analyzing data and near-real-time tracking of objects such as troop or refugee movements along borders in violent conflict zones.[286][288]

Artificial intelligence can be used to mitigate vital cross-national diplomatic talks to prevent translation errors caused by human translators.[289] A major example is the 2021 Anchorage meetings held between US and China aimed at stabilizing foreign relations, only for it to have the opposite effect, increasing tension and aggressiveness between the two nations, due to translation errors caused by human translators.[290] In the meeting, when United States National Security Advisor to President Joe Biden, Jacob Jeremiah Sullivan stated, "We do not seek conflict, but we welcome stiff competition and we will always stand up for our principles, for our people, and for our friends", it was mistranslated into Chinese as "we will face competition between us, and will present our stance in a very clear manner", adding an aggressive tone to the speech.[290] AI's ability for fast and efficient natural language processing and real-time translation and transliteration makes it an important tool for foreign-policy communication between nations and prevents unintended mistranslation.[291]

These were the four the most widely used AI textbooks in 2008: