Alan Matheson Turing was a British mathematician and scientist who in 1936 proposed the idea of the “Universal Machine” (Biography, 2014). The “Universal Machine” or “Turning Machine” was an early conceptualization of the current computer (Biography, 2014). In the 1950s, Turing proposed and initiated the “Turing Test” to test the intelligence of machines (Biography, 2014). Turing was open minded in his view of intelligence and believed that as technology improved, it would become evident that machines can be programmed to think and be designed to mimic human thoughts and actions (Turing, 1950).
John McCarthy was an American mathematician and Professor of Computer Science at Stanford University from 1962-2000 (“John McCarthy”, 2022). McCarthy is considered to be one of the “founding fathers” of artificial intelligence, as he along with four other scholars invented the term in 1956 (“John McCarthy”, 2022). McCarthy believed that Artificial Intelligence was the process of programming machines that would be capable of completing tasks and achieving goals (McCarthy, 2007).
Herbert A. Simon was an American political scientist with multi-disciplinary interests in the fields of computer science, economics, and cognitive psychology (“Herbert A. Simon”, 2022). Simon is also considered to be one of the “founding fathers” of artificial intelligence (UBS, n.d.). Simon was a visionary who believed that artificial intelligence would result in machines that would eventually be able to do any task that humans are able to do (UBS, n.d.). Simon developed the theory that machines could be programmed to simulate human problem solving (Newell & Simon, 1961).
Marvin Minsky was an American mathematician and computer scientist who co-founded MIT’s Artificial Intelligence laboratory in 1959 (BBC, 2016). Minsky’s view on artificial intelligence was that the human mind would eventually be replicated in a machine, including the ability to use reason in decision making (BBC, 2016).
Timnit Gebru is an Ethiopian-American computer scientist who specializes in topics related to the bias of algorithms and data mining (“Timnit Gebru”, 2022). Gebru previously worked in the field of AI ethics for Google, where Gebru and some colleagues attempted to sound the alarm of the risks involved in the ability of artificial intelligence to produce its own text based on its learning from large language models (Hao, 2020). According to Gebru, AI is unable to understand the main ideas and develop meaning from research, but is capable of synthesizing large amounts of data and producing intelligible text that contributes to bias and misinformation (Hao, 2020).
How do “machine (programming) languages” differ from human (natural) ones?
The main goal of all languages is communication (Harris, 2018). However, programming languages are designed for the purpose of communicating instructions to a machine, whereas human languages are for communicating thoughts, ideas, and emotions (Harris, 2018). Programming languages are specific and require logic for the machine to understand the instructions (Harris, 2018). On the other hand, human languages are full of nuance and are continually evolving (Harris, 2018).
How does “machine (artificial) intelligence” differ from the human version?
The tendency has been to look at a machine’s ability to complete a specific skill when measuring machine intelligence (Chollet, 2019). Chollet (2019) reminds us that human intelligence consists of more than the ability to acquire and execute skills, but that “the hallmark of broad abilities (including general intelligence) is the power to adapt to change, acquire skills, and solve previously unseen problems – not skill itself, which is merely the crystallized output of the process of intelligence” (p. 20). Thus, human intelligence is based upon a foundation of knowledge and experiences that allows humans to transfer that knowledge to different scenarios, whereas machine intelligence is how skillfully a machine can complete a task (Chollet, 2019).
How does “machine learning” differ from human learning?
Machine learning is different from human learning in that humans build their knowledge and understanding from a variety of different sources and experiences, whereas machines learn from the data they are exposed to from their developer (Heilweil, 2020). The problem of learning solely from data is that while both humans and machines can develop bias in their learning, humans can learn to check for bias in the information they are exposed to and unlearn biased beliefs or behaviours whereas machines cannot. In fact, machines provide the illusion of being neutral in their learning and behaviour, when the reality is that their bias is hidden and the impact of their bias may be unknown but incredibly damaging (Heilweil, 2020).
Turing Test: How does a machine’s response differ from a human’s response?
When answering the above questions, a human is better at being able to evaluate whether the information is relevant and appropriate for the audience. A human author is also better able to connect the information to their existing knowledge and experiences, and perhaps provide relevant, engaging examples or anecdotes. A machine would respond differently than a human to the above questions, similar to how Google Nest or Siri respond when they are asked a question. Machines tend to identify the most popular source of information on the topic and use this information to answer the question. Thus, a machine would be able to identify some of the more important pieces of information, but it might include additional information that is less relevant and does not belong due to the fact that the algorithm somehow connected the two topics. For example, when discussing AI, the machine might bring up information from a movie or work of science fiction, as it would not be able to distinguish that these are not reliable sources of information. Alternatively, the machine may omit important information for unknown reasons or biases related to its algorithm, maybe favoring articles written by men over women.
References
BBC News. (2016, January 26). AI pioneer Marvin Minsky dies age 88. https://www.bbc.com/news/technology-35409119
Biography. (2014, October 30). Alan Turing Biography. A & E Television Networks. https://www.biography.com/scientist/alan-turing
Buolamwini, J. (2019, February 7). Artificial intelligence has a problem with gender and racial bias. Here’s how to solve it. Time. https://time.com/5520558/artificial-intelligence-racial-gender-bias/
Chollet, F. (2019, November 5). On the measure of intelligence. arXiv: 1911.01547. https://arxiv.org/pdf/1911.01547.pdf
Harris, A. (2018, October 31). Human languages vs. programming languages. Medium. https://medium.com/@anaharris/human-languages-vs-programming-languages-c89410f13252
Heilweil, R. (2020, February 18). Why algorithms can be racist and sexist. Vox. https://www.vox.com/recode/2020/2/18/21121286/algorithms-bias-discrimination-facial-recognition-transparency
Herbert A. Simon. (2022, February 19). In Wikipedia. https://en.wikipedia.org/w/index.php?title=Herbert_A._Simon&oldid=1072765316
Hao, K. (2020, December 4). We read the paper that forced Timnit Gebru out of Google. Here’s what it says. MIT Technology Review. https://www.technologyreview.com/2020/12/04/1013294/google-ai-ethics-research-paper-forced-out-timnit-gebru
John McCarthy (computer scientist). (2022, February 21), In Wikipedia. https://en.wikipedia.org/w/index.php?title=John_McCarthy_(computer_scientist)&oldid=1073094315
McCarthy, J. (2007). Basic questions. http://www-formal.stanford.edu/jmc/whatisai/node1.html
Newell, A., & Simon, H. A. (1961). Computer simulation of human thinking. Science, 134(3495), 2011–2017. http://www.jstor.org/stable/1708146
Timnit Gebru. (2022, February 20). In Wikipedia. https://en.wikipedia.org/w/index.php?title=Timnit_Gebru&oldid=1072924053
Turing, A.M. (1950). Computing Machinery and Intelligence. Mind 49: 433-460.
UBS. (n.d.). Herbert A. Simon. Nobel Perspectives. https://www.ubs.com/microsites/nobel-perspectives/en/laureates/herbert-simon.html