1.1 Defining the terms
Within academic and scientific spheres, the concepts surrounding artificial intelligence are often confined to an understanding of intelligence that is largely computational in nature and scope. In its most basic form, AI can be understood as an intentional action aimed at producing a certain outcome. Sometimes the input/output data is known, sometimes unknown – at least until the processing is complete. Yet the tendency remains to evaluate AI largely in relation to human intelligence, which goes beyond mere electronic processing and considers wider, sometimes unexpected consequences. Margaret Boden, Professor of Cognitive Science and Informatics at the University of Sussex, argues that:
Artificial Intelligence seeks to make computers do the sorts of things that minds can do … Intelligence isn’t a single dimension, but a richly structured space of diverse information-processing capacities. Accordingly, AI uses many different techniques, addressing many different tasks.[1]
However, viewing AI in relation to human intelligence presents significant challenges that much of the literature has struggled to explain or classify. For instance, if AI is to mimic human intelligence, how does this relate to conscience? Unlike the electronic calculator, human intelligence performs a simultaneous array of calculations alongside the given task. Again, some of these calculations are known, others are not. This raises challenges by making human intelligence less predictable, less quantifiable and consequently harder to emulate within traditional AI systems.
Since the turn of the century a distinction has been growing between narrow AI and general AI, or artificial general intelligence (AGI).[2] Narrow AI is the simple, single-task operation conducted by AI. It follows a concrete set of instructions and usually gives a predictable result. Examples include computer games, simple data analysis, certain algorithms – such as email filtering – and other applications encountered in day-to-day life yet often remaining hidden. Narrow AI ‘displays a certain degree of intelligence in a particular field. It performs highly specialized tasks for humans, within that narrow field.’[3] More recently, narrow AI has evolved to encompass learning abilities, known as machine learning or deep learning. This involves not just executing a set task or goal but also learning from new data and adapting the execution based on input data – not a predetermined computation.
Artificial general intelligence, however, remains a hypothesis. It is a far more ambitious project that seeks to replicate the human brain in the entirety of its complexity, and potentially surpass it. The distinction between narrow AI and AGI is significant because the latter has intrinsic agency to act, a capacity to learn and an ability to perform actions based on the data it generates. Colloquially speaking, the machine has a mind of its own. Thinkers such as Yuval Noah Harari and Jaan Tallinn are proponents of this apocalyptic existential threat of an AI ‘singularity’ point in time, whereby the machine has improved itself to such an extent that it surpasses human intelligence and overpowers the human race.[4]
Yet there is a crucial shortcoming: unlike the human mind, AGI fundamentally lacks non-computational thinking – or conscience – and a deeper sense of self-awareness that goes beyond algorithmic learning. By default, therefore, AGI cannot possess any inherent gauge of morality except for general rules either preprogrammed or learned, which raises key existential questions: Is sentience something that can be programmed or does it originate from an external source? If, hypothetically speaking, consciousness can be programmed, would it be genuine consciousness (in a human sense)? Indeed, what is genuine consciousness? Would it even matter as long as there are positive and ‘moral’ outcomes? John Lennox, Professor of Mathematics at Oxford University, has this to say about genuine sentience:
it is one thing to make a machine that can simulate, say, a human hand lifting an object; it is a completely different thing to make a machine that can simulate the thoughts of a human when he or she is lifting an object. It is much easier to do the first than the second, and if utility is all that is required, then the first is all that is necessary … For AI computer systems have no conscience, and so the morality of any decisions they make will reflect the morality of the computer programmers – and that is where the difficulties start.[5]
This lack of embedded morality or the ability for self-reflection against a moral code is indeed where the difficulties start for complex systems such as AGI. Could an AGI compose a poem about war? It most likely could – and in some sense current, generative AI already has. The Appendix offers two poems generated by ChatGPT, an artificial intelligence chatbot developed by OpenAI and launched in November 2022. The two poems are on the topics of consciousness and war and make for intriguing reading.
However, at what level would a superintelligent AI – or ChatGPT currently – engage with the topic of war? Indeed, can feelings and emotions about the gravity of war be manually implemented or, in the case of machine learning, can an AGI be encouraged to learn feelings of sorrow or loss? If it does express feelings of sorrow or loss, would these be anything more than computational? Would it even matter as long as humans are unable to distinguish between the two? Chapter 3 returns to some of these questions, discussing this interrelationship in greater depth.
Since the 1980s, AI’s utility has moved from being its principal objective to one of many. Creativity and innovation produced by AI itself are now held in higher regard and seen as more enticing to programmers than utility alone. Professor Stuart J. Russell points out that historically there are two different approaches to AI.[6] One is viewing AI in ‘fidelity to human performance’, the other is a more abstract ‘definition [of intelligence] called rationality – [or] loosely speaking, doing the “right thing”’.[7] The two therefore give rise respectively to artificial systems that aim to mimic humans in both behaviour and output, and those primarily rational in behaviour and output. These differing approaches lead to rather different pursuits of recreating intelligence, whether akin to human intelligence or not – demonstrating the breadth and depth of the issue.
Dr John McCarthy from Stanford University defines AI as:
the science and engineering of making intelligent machines, especially intelligent computer programs. It is related to the similar task of using computers to understand human intelligence, but AI does not have to confine itself to methods that are biologically observable.[8]
He argues that there is an entire list of branches within the field of AI, many not yet clearly defined or even apprehended. Areas of enquiry include, inter alia, logical AI, pattern recognition, representation, inference, heuristics and genetic programming.[9]
Attempting to define AI brings its own set of challenges, not least because the continually evolving space results in an ever-growing number of known unknowns, as well as unknown unknowns, or future pathways of AI evolution that scientists have not yet thought about. It also brings further challenges because the distinction between narrow AI and AGI does not give sufficient weight to the sheer complexity and difficulty of the latter. To create something that can rival the human brain, not just in utility but also in a sense of conscience, situational awareness and the biochemical-emotional relationship, remains no more than an aspiration. For now, we can define artificial intelligence as the leveraging of computer power – in all forms – to computational tasks such as learning, problem-solving and data generation.
1.2 Milestones in AI history
The idea of creating intelligent machines dates back to antiquity. For millennia, humanity has been intrigued – in some cases obsessed – with the idea of creating objects that to a greater or lesser degree reflect their own image. The word ‘automation’ comes from the Greek automatos, meaning ‘self-acting, moving or acting on its own’.[10] It was coined by the Greek-Egyptian mathematician and inventor Heron of Alexandria (c.AD 0–70), when he created the first set of automated temple doors. Driven by a steam-powered hydraulic mechanism, the doors would open when the fire to the temple altar was lit and close when it was extinguished – much to the astonishment of devotees.[11] In his writings and early sketches Heron named the invention ‘Number 37’, marking an important milestone because it represents – albeit this is difficult to prove conclusively – the earliest form of a human-engineered automated machine. Number 37 was the first mechanism with the ability to act on its own.
Other objects considered at the time to have some degree of intrinsic intelligence were principally objects of divine worship. Their capabilities were wholly or in part derived from their divine status. Belief in them required faith, not observation. The Introduction mentioned Thalos, a giant automaton of bronze built by Hephaestus, the Greek mythological God of invention and blacksmithing. Thalos was ordered by Zeus, king of the Gods, to patrol Crete three times a day to protect it from foreign invaders. At the core of his body Thalos held his mysterious life source, something the Greeks called ‘ichor’. Mythology literature reveals how Thalos was killed by a sorcerer who pulled a bolt from his ankle, which caused the ichor fluid to flow out.[12] Yet Thalos was a myth, and his existence in the imagination of the ancient Greeks illustrates the powerful allure of machines that embody superior human features. For Greeks, Thalos represented what they lacked themselves: a large and powerful human-like machine that reaped the benefits of technological advancement and was blessed by the gods – in this case Zeus – to offer protection from enemies.
Leaving aside objects such as figurines or statues given divine attributes, the abacus is often regarded as the first intelligent machine from a computational standpoint. The earliest archaeological evidence of the abacus dates back to the third millennium BC, within the Babylonian empire and the Early Dynastic period in Egypt. Ettore Carruccio, a scholar of the Babylonian period, pointed to evidence that Babylonians ‘may have used the abacus for the operations of addition and subtraction’.[13] Though simple by contemporary standards, the abacus was probably the first tool that enabled a visual representation of basic calculations, thus reducing the possibility for error and expanding understanding of elementary mathematics to a wider audience. Its impact was vast and to some extent the abacus can be viewed as a primitive form of personal computing device.
Moving briskly forward to the early 1800s, we arrive at the British mathematician, philosopher and inventor Charles Babbage, who proposed the Analytical Engine. Although Babbage himself never had the opportunity to complete the build, his design marked a milestone in the field of computer science – he was a pioneer in the development of intelligent machines. His Analytical Engine was a much-improved successor to his previous design, the Difference Engine, and it achieved what was later known as ‘turning complete’. This meant that it incorporated the same basic logical structures that electronic computers would later possess. The Analytical Engine comprised of an arithmetic logic unit, ‘control flow’ in the form of conditional branching and loops, and integrated memory – making it the first general-purpose programmable computing engine.[14] Babbage became known as the father of computing for his remarkable skill in precision engineering, creativity of design and intellectual curiosity.[15] This paved the way for computer science to develop not just in its practical applications – which were themselves useful even in the early days – but as a recognised academic field of inquiry. Machines such as the Difference Engine were mostly used for applications that required tabulating polynomial functions. Despite Babbage’s remarkable contributions to the field, he failed to secure sufficient funding to support his work and bring many of his projects to completion which, according to one source, ‘left Babbage in his declining years a disappointed and embittered man’.[16]
Another key figure in the history of AI was Alan Turing. If Charles Babbage was the father of computing, Turing took that role for modern computer science.[17] Without going into detail on Turing’s many accomplishments, it is worth mentioning some of his key contributions. ‘The genius of Bletchley Park’, Alan Turing is best known in popular culture for cracking the German Enigma code during the Second World War. It is thought that Churchill once credited him with making the single biggest contribution to the Allied victory against Nazi Germany. President Eisenhower noted in a letter that the work at Bletchley Park ‘saved countless British and American lives and, in no small way, contributed to the speed with which the enemy was routed and eventually forced to surrender’.[18]
But concerning artificial intelligence, Turing produced two seminal papers that effectively established the field of computer science. What was novel about his thought was that by as early as 1937 he imagined a machine that could be instructed by multiple computations with consequentially varied outcomes. This in turn could be used to resolve a diverse array of computational problems. It was in some sense the earliest feasible description of a multipurpose and multitasking computer. At a time when machines were limited to singular preprogrammed computations, Turing’s first paper – ‘On Computable Numbers, with an Application to the Entscheidungsproblem’ – was revolutionary.[19] Professor Joel Hamkins argues that:
The paper is an incredible achievement. He accomplishes so much: he defines and explains the machines; he proves that there is a universal Turing machine … he argues that his machine concept captures our intuitive notion of computability; and he develops the theory of computable real numbers.[20]
In his mathematical descriptions Turing called it the Universal Machine, though it would later become more commonly known as the Turing Machine. What makes the machine special is that it could, in theory, solve any mathematical problem presented to it in symbolic form. This is largely because Turing developed it along similar lines to a human carrying out mathematical computational processes. Yet some computer scientists and mathematicians, such as Professor Martin Davis, argue that the true legacy of Turing’s Universal Machine is the development of the ‘the stored-program computer’: placing the input data as well as the machine instructions on a uniform memory, something present today in all computer devices.[21] Paul Gray from Time magazine wrote that ‘everyone who taps at a keyboard … is working on an incarnation of a Turing machine.’[22]
The second Turing paper to touch on here was ‘Computing Machinery and Intelligence’, published in 1950.[23] It postulates the question of machine intelligence – Can machines think? – and attempts to demonstrate intelligence via an imitation game: Can AI become indistinguishable from human intelligence and can it fool a human into thinking it is human itself? This would commonly become known as the Turing test.
Illustrated in Figure 1, the Turing test involves three actors: two guests and an interrogator. One of the guests is the AI (person A) while the other is a human (person B). The AI has to trick the interrogator (person C, also human) into thinking it is human. The ‘conversation’ has a duration of five minutes and if the AI successfully convinces the interrogator it is human over 30 per cent of the time, the test has been passed.
The interview consists of a series of questions and answers whereby the interrogator probes A and B individually. One of the major attractions of the test is its simplicity and flexibility: it allows the interrogator to pose a wide variety of questions and intellectual – or emotional – challenges. Turing himself wrote that: ‘The question and answer method seems to be suitable for introducing almost any one of the fields of human endeavour that we wish to include.’[24]
The test consequently had a significant impact not just on the development of artificial intelligence itself but on how we measure AI in relation to human intelligence. It sparked the debate on our capacity – or incapacity – to quantify and evaluate intelligence, itself an abstract, relative term. The test therefore represents more of a conceptual exercise than a scientific measuring tool. Professor Daniel Dennett has said that Alan Turing intended it as ‘a conversation stopper, as a thought experiment that should convince people that any computer that would pass this test – of course it would be intelligent’.[25]
One important challenge of the Turing test is its reliance on humans as judges in classifying the subject in question. This has caused some unreliable results because humans can be poor at recognising other humans through text alone. They often fall prey to what is known as the Eliza effect; that is, our subconscious tendency to view computer intelligence as analogous to human intelligence – a form of anthropomorphisation (in this case, of computer systems).
A good example of the anthropomorphic Eliza effect is given by the computer scientist Douglas Hofstadter, and involves a cash machine that displays the words ‘Thank You’ at the end of a transaction. The human may consciously or subconsciously think that it is expressing gratitude when all the machine is doing is showing a preprogrammed automated response.[26] The Eliza effect can thus lead to some unintelligent machines passing the test. For instance, delaying the response or remaining silent can be misinterpreted as thought processing. There have even been cases in which humans were misidentified as machines – known as the Confederate effect.[27] Here Kevin Warwick and Huma Shah have conducted research and found that, rather counterintuitively, under testing conditions humans attempting to ‘act human’ actually appear less ‘human-like’ and increasingly prone to giving computer-type responses.[28]
1.3 AI in the information age (1950s onwards)
To talk about the history of AI in the information age – or technological era – is, in effect, to discuss the history of computer science. The term ‘artificial intelligence’ was only coined during the 1956 Dartmouth College conference in Hanover, New Hampshire.[29] Following Turing’s seminal work, the 1950s and 60s marked a period of great enthusiasm for machine intelligence. Turing himself believed that human-level artificial intelligence would be achieved within 50 years.[30] The MIT cognitive scientist Marvin Minsky, and others who attended the Dartmouth conference, were extremely optimistic about AI’s future: within ‘a generation … the problem of creating “artificial intelligence” will substantially be solved’.[31] Scientists focused on challenges – mathematical or in the form of logic puzzles, games, pattern interpretation – that mimicked or were considered indicative of human intelligence, and attempted to meet them with AI.
Curiously, the personal computer (PC), when first invented, was not recognised as a milestone in the development of AI. In 1974 a small electronics company called MITS created the Altair 8800, the first PC to earn widespread popularity among the early tech community. Ed Roberts, the founder of MITS, claimed that as many as 40,000 units were sold, although the precise number remains uncertain.[32]
What the Altair 8800 did was spark a revolution in demand for personal computing. This came at a time when many doubts were cast over the purpose and usefulness of PCs within the household. ‘Computerphobia’, as it was known in the 1970s and 80s, became a real issue: ‘computers won’t make toast or vacuum a carpet’, wrote the journalist Charles Rubin in 1983.[33] Early research suggested that fuelling this phobia were fears, among others, of damaging the computer, looking foolish, losing electricity and lacking control over the machine.[34] There was also a period in the early stages when large swathes of the public, as well as some experts, struggled to find much use for the PC. The computer scientist Ken Olsen famously said in 1977: ‘There is no reason for any individual to have a computer in his home’ (although Olsen was speaking in reference to computers being used for home automation).[35] Adoption of the PC and, by extension, the development of AI, therefore proved by no means straightforward, and slow progress as well as lack of funding led to the period 1974–80 becoming known as the first ‘AI winter’.[36]
Despite these hurdles, the early 1980s were marked by a rebound in the development of new hardware as well as more intuitive, user-friendly software. The Xerox 8010 Star, released in 1981, was the first commercial personal computer to incorporate many of the core technologies that PCs feature today, including a graphical user interface, an operating system with icons and folders, a two-button mouse, Ethernet connectivity and even email.[37]
The Star’s graphical interface became known as What You See Is What You Get (WYSIWYG), where users were encouraged to take actions directly on screen via the mouse and keyboard. For the first time, visual elements such as icons, taskbars, folders and documents represented interactive objects. WYSIWYG was revolutionary not just because it implemented what was then considered a seamless interface but because it offered the new possibility for those without prior knowledge of coding to use and interact with the device. Many of the functions and abilities of the computer became self-taught through simple trial and error. The intuitive design language encouraged users to think of the icons on the screen the same way they would their physical counterparts; that is, folders, trash, pictures, games and so on.[38] At the time, a young Steve Jobs was offered an opportunity to visit Xerox’s Palo Alto Research Centre. Recalling the trip, Jobs said Xerox:
showed me three things, but I was so blinded by the first one that I didn’t really see the other two … The graphical user interface, I thought, was the best thing I’d seen in my life … Within 10 minutes it was obvious to me that all computers would work like this, someday. It was obvious.[39]
Steve Jobs’s Apple Macintosh went on to sell a record 4.5 million computers in 1995 alone.[40]
1.4 Current trends in AI
The distinction between narrow AI and AGI was considered above, noting how narrow AI represents simple, often single-task computation that follows a concrete set of instructions and usually gives a predictable result. Much of the discussion so far has been of narrow AI, but now turns to an expansion of it that first emerged in the 1980s known as machine learning (ML).
Dr Issam Naqa defines ML as ‘an evolving branch of computational algorithms that are designed to emulate human intelligence by learning from the surrounding environment’.[41]IBM sees it as ‘a branch of artificial intelligence (AI) and computer science which focuses on the use of data and algorithms to imitate the way that humans learn, gradually improving its accuracy’.[42] Generally accepted definitions of ML bear certain similarities. Therefore a working definition that encompasses the fundamentals might say that ML is a branch within AI where computer systems have the ability to learn from new data and perform specific tasks with it. This data can be acquired from external sources or generated by the program itself, as would be the case in generative AI, for example.
One notable advancement in the field of ML learning came in 1981 when Professor Gerald DeJong proposed explanation-based learning (EBL). Although DeJong himself admits that ‘there is yet no satisfactory answer to the question’ of what exactly EBL is, we can broadly see it as a strategy of ML that learns from examples and has the capacity to draw generalisations or form concepts.[43] Stuart Russell defines explanation-based learning as ‘a method for extracting general rules from individual observations’.[44] For instance, EBL has the capacity to create a program that learns to play chess by taking training examples and determining what are the relevant features in order to form a generalisation. EBL is usually based on four inputs: a hypothesis, a domain theory, training examples and the operationality criteria (determining which features are efficiently recognisable).[45] It is important because it represents one of the earliest forms of machine learning that attempts to imitate the human process of learning; that is, developing conclusions through the gathering of information and utilisation of prior examples.
Deep learning is a further evolution of machine learning that has dominated discourse in more recent times. If machine learning represents the ability to learn without explicit programming, deep learning takes matters a step further and utilises the gathered data to extract conclusions via neural networks. In this sense, it represents a more complex subset method of machine learning. Yet the term itself encompasses a broader variety of techniques, which are centred around algebraic circuits – ‘deep’ refers to the fact that in many cases the circuits have multiple layers (akin to a human neural network), making the computational path from input to output in many cases quite extensive.[46]
Figure 2 illustrates a basic example of how deep learning works in practice. Suppose an AI can differentiate between what is and is not a car. Under machine learning the features of what make up a car are manually programmed into the algorithm – this is known as ‘feature extraction’. Under deep learning the artificial neural network picks up the distinct array of features itself without the need of manual input: it does not require predetermined ‘labelled data’ to draw conclusions. However, this increased degree of autonomy demands a greater volume of data. Thus the AI is able to learn the characteristic features of a car through the automatic processing of an extensive number of examples.
From a functional standpoint, machine learning in general and deep learning in particular carry fundamental implications for how we interact with our work, social activities, families and even ourselves. Algorithmic calculations dictate what we see and often what we do not see in the virtual sphere. Yet machine learning is all the more pertinent precisely because new ML technologies such as generative AI are becoming increasingly embedded features of contemporary life.
This chapter has briefly sketched the history and some of the key milestones in the development of AI, as a basis for considering the complex implications of its utilisation and adoption. Chapter 2 takes a more practical approach by considering two case studies in AI.
Notes to Chapter 1
[1] Margaret Boden, Artificial Intelligence: A Very Brief Introduction, Oxford: Oxford University Press, 2018, p. 1.
[2] ‘Narrow AI vs. General AI – What’s Next for the Future of Tech?’, Analytics Insight, 5 February 2022, https://www.analyticsinsight.net/narrow-ai-vs-general-ai-whats-next-for-the-future-of-tech/.
[3] ‘Narrow AI vs. General AI’, Analytics Insight.
[4] Benjamin Charles Germain Lee, ‘The Singularity Prophets’, Current Affairs, 23 July 2020, https://www.currentaffairs.org/2020/07/the-singularity-prophets.
[5] John C. Lennox, 2084: Artificial Intelligence and the Future of Humanity, Grand Rapids, MI: Zondervan, 2020, pp. 17, 145.
[6] Stuart Russell and Peter Norvig, Artificial Intelligence: A Modern Approach, Global edition, Harlow: Pearson, 2021, pp. 19–20.
[7] Russell and Norvig, Artificial Intelligence, pp. 19–20.
[8] Richard S. Sutton, ‘John McCarthy’s Definition of Intelligence’, Journal of Artificial General Intelligence 11:2 (2020), pp. 66–7.
[9] Sutton, ‘John McCarthy’s Definition of Intelligence’, pp. 66–7.
[10] Online Etymology Dictionary, ‘automatic’, 2023, https://www.etymonline.com/word/automatic.
[11] ‘Heron of Alexandria – Automated Temple Doors’, Artefacts – Scientific Illustration and Archaeological Reconstruction, 2021, http://www.artefacts-berlin.de/portfolio-item/heron-of-alexandria-automated-temple-doors/.
[12] ‘Stanford Researcher examines Earliest Concepts of Artificial Intelligence, Robots in Ancient Myths’, Stanford News, 28 February 2019, https://news.stanford.edu/2019/02/28/ancient-myths-reveal-early-fantasies-artificial-life/.
[13] Ettore Carruccio, Mathematics and Logic in History and in Contemporary Thought, trans. Isabel Quigly, Piscataway, NJ: Aldine Transaction, 2006, p. 14.
[14] ‘The Engines’, Computer History Museum, 15 January 2023, https://www.computerhistory.org/babbage/engines.
[15] University of Minnesota, Charles Babbage Institute, ‘Computing, Information and Culture: Who was Charles Babbage?’, College of Science and Engineering and University Libraries, https://cse.umn.edu/cbi/who-was-charles-babbage.
[16] University of Minnesota, Charles Babbage Institute, ‘Computing, Information and Culture’.
[17] ‘Alan Turing: The Father of Modern Computer Science’, The Twickenham Museum, 2023 http://www.twickenham-museum.org.uk/detail.php?aid=66&ctid=1&cid=13.
[18] Patrick Sawer, ‘Letter reveals Bletchley Park Code Breakers secretly thanked by General Eisenhower for “Priceless” Work’, Eisenhower Foundation, 15 March 2016, https://www.eisenhowerfoundation.net/sites/default/files/2020-05/File%206%2C%20Article%201.pdf.
[19] Alan Turing, ‘On Computable Numbers, with an Application to the Entscheidungsproblem’, Proceedings of the London Mathematical Society Series 2, 42:1 (1937), pp. 230–65.
[20] Joel David Hamkins, ‘Alan Turing, On Computable Numbers’, Joel David Hamkins, 29 September 2018, http://jdh.hamkins.org/alan-turing-on-computable-numbers/.
[21] Martin Davis, The Universal Computer: The Road from Leibniz to Turing, New York: Norton, 2000, p. 121.
[22] Paul Gray, ‘Computer Scientist – Alan Turing: While addressing a Problem in the Arcane Field of Mathematical Logic, he imagined a Machine that could mimic Human Reasoning. Sound familiar?’ Time Magazine Archives, 29 March 1999, https://content.time.com/time/subscriber/article/0,33009,990624-3,00.html.
[23] Alan Turing, ‘Computing Machinery and Intelligence’, Mind 59:236 (1950), pp. 433–60.
[24] Turing, ‘Computing Machinery and Intelligence’.
[25] ‘Turing Test – Daniel Dennett’, YouTube, 22 May 2022, https://www.youtube.com/watch?v=e8vZy8a9lSc.
[26] Douglas R. Hofstadter and the Fluid Analysis Research Group, ‘Preface 4: The Ineradicable Eliza Effect and its Dangers, Epilogue’, in Fluid Concepts and Creative Analogies: Computer Models of the Fundamental Mechanisms of Thought, New York: Basic Books, 1996, p. 157.
[27] Huma Shah and Odette Henry, ‘The Confederate Effect in Human–Machine Textual Interaction’, conference paper, 2005.
[28] Kevin Warwick and Huma Shah, ‘Human Misidentification in Turing Tests’, Journal of Experimental and Theoretical Artificial Intelligence 27:2 (2015), pp. 123–35.
[29] Jørgen Veisdal, ‘The Birthplace of AI: The 1956 Dartmouth Workshop’, Medium, 12 September 2019, https://www.cantorsparadise.com/the-birthplace-of-ai-9ab7d4e5fb00.
[30] H. P. Newquist, The Brain Makers: The History of Artificial Intelligence – Genius, Ego, and Greed in the Quest for Machines that Think, New York: The Relayer Group, 2020, p. 51.
[31] Daniel Crevier, AI: The Tumultuous Search for Artificial Intelligence, New York: Basic Books, 1993.
[32] Adwater & Stir, ‘The First?’, 15 February 2023, https://adwaterandstir.com/altair-history/.
[33] David Neil Crocker, ‘Excerpts from “Some People should be afraid of Computers”, but Modernized’, David Neil Crocker, 26 March 2019, http://crockedile.com/2019/03/26/textual-poaching-computerphobia/.
[34] Adrienne LaFrance, ‘When People Feared Computers’, The Atlantic, 30 March 2015, https://www.theatlantic.com/technology/archive/2015/03/when-people-feared-computers/388919/.
[35] Alex Bracetti, ‘The Spam Crisis Solved? The 25 Craziest Things ever said by Tech CEOs’, Complex, 14 January 2013, https://www.complex.com/pop-culture/2013/01/the-25-craziest-things-ever-said-by-tech-ceos.
[36] Jim Howe, ‘Artificial Intelligence at Edinburgh University: A Perspective’, 2007, https://www.inf.ed.ac.uk/about/AIhistory.html.
[37] ‘Xerox Star 8010 Information System, 1981’, The Interface Experience: 40 Years of Personal Computing (Bard Graduate Center), 15 January 2023, https://interface-experience.org/objects/xerox-star-8010-information-system/.
[38] ‘Xerox Star 8010 Information System, 1981’.
[39] ‘How Steve Jobs got the Ideas of GUI from XEROX’, YouTube, 4 January 2014, https://www.youtube.com/watch?v=J33pVRdxWbw.
[40] Jeremy Reimer, ‘Total Share: 30 Years of Personal Computer Market Share Figures – It’s been a Long, Strange Trip for the Personal Computer over 30 Years’, Arstechnica, 15 December 2005, https://arstechnica.com/features/2005/12/total-share/7/.
[41] Issam El Naqa, Ruijiang Li and Martin J. Murphy, 2015, ‘What is Machine Learning?’, in El Naqa, Li and Murphy (eds), Machine Learning in Radiation Oncology: Theory and Applications, Cham: Springer, 2021, https://link.springer.com/chapter/10.1007/978-3-319-18305-3_1.
[42] ‘Machine Learning: What is Machine Learning?’, IBM Cloud Learn Hub, 11 January 2023, https://www.ibm.com/cloud/learn/machine-learning.
[43] Howard E. Shrobe, Exploring Artificial Intelligence: Survey Talks from the National Conferences on Artificial Intelligence, Amsterdam: Elsevier Science, 2014.
[44] Russell and Norvig, Artificial Intelligence, p. 750.
[45] Richard M. Keller, ‘Defining Operationality for Explanation-Based Learning’, Artificial Intelligence 35:2 (1988), pp. 227–41.
[46] Russell and Norvig, Artificial Intelligence, p. 801.