Editing Artificial general intelligence (section)

== Characteristics ==
{{Main|Artificial intelligence}}

Various popular definitions of [[intelligence]] have been proposed. One of the leading proposals is the [[Turing test]]. However, there are other well-known definitions, and some researchers disagree with the more popular approaches.{{Efn|AI founder [[John McCarthy (computer scientist)|John McCarthy]] writes: "we cannot yet characterize in general what kinds of computational procedures we want to call intelligent."<ref>{{Cite web |last=McCarthy |first=John |author-link=John McCarthy (computer scientist) |date=2007a |title=Basic Questions |url=http://www-formal.stanford.edu/jmc/whatisai/node1.html |url-status=live |archive-url=https://web.archive.org/web/20071026100601/http://www-formal.stanford.edu/jmc/whatisai/node1.html |archive-date=26 October 2007 |access-date=6 December 2007 |publisher=[[Stanford University]]}}</ref> (For a discussion of some definitions of intelligence used by [[artificial intelligence]] researchers, see [[philosophy of artificial intelligence]].)}}

=== Intelligence traits ===
Researchers generally hold that a system is required to do all of the following to be regarded as an AGI:<ref name=":12">This list of intelligent traits is based on the topics covered by major AI textbooks, including: {{Harvnb|Russell|Norvig|2003}}, {{Harvnb|Luger|Stubblefield|2004}}, {{Harvnb|Poole|Mackworth|Goebel|1998}} and {{Harvnb|Nilsson|1998}}.</ref>
* [[automated reasoning|reason]], use strategy, solve puzzles, and make judgments under [[uncertainty]]
* [[knowledge representation|represent knowledge]], including [[commonsense knowledge (artificial intelligence)|common sense knowledge]]
* [[automated planning and scheduling|plan]]
* [[machine learning|learn]]
* communicate in [[natural language processing|natural language]]
* if necessary, [[Artificial intelligence systems integration|integrate these skills]] in completion of any given goal
Many [[Interdisciplinarity|interdisciplinary]] approaches (e.g. [[cognitive science]], [[computational intelligence]], and [[decision making]]) consider additional traits such as [[imagination]] (the ability to form novel mental images and concepts)<ref>{{Harvnb|Johnson|1987}}</ref> and [[Self-determination theory|autonomy]].<ref>de Charms, R. (1968). Personal causation. New York: Academic Press.</ref>

Computer-based systems that exhibit many of these capabilities exist (e.g. see [[computational creativity]], [[automated reasoning]], [[decision support system]], [[robot]], [[evolutionary computation]], [[intelligent agent]]). There is debate about whether modern AI systems possess them to an adequate degree.<ref>{{cite journal|last=Van Eyghen|first= Hans|title=AI Algorithms as (Un)virtuous Knowers|journal=Discover Artificial Intelligence|volume=5|issue=2|date=2025|doi= 10.1007/s44163-024-00219-z|doi-access=free}}</ref>

=== Physical traits ===
Other capabilities are considered desirable in intelligent systems, as they may affect intelligence or aid in its expression. These include:<ref name=":13">Pfeifer, R. and Bongard J. C., How the body shapes the way we think: a new view of intelligence (The MIT Press, 2007). {{ISBN|0-2621-6239-3}}</ref>
* the ability to [[machine perception|sense]] (e.g. [[computer vision|see]], hear, etc.), and
* the ability to act (e.g. [[robotics|move and manipulate objects]], change location to explore, etc.)
This includes the ability to detect and respond to [[hazard]].<ref name="White 1959 297–333">{{Cite journal |last=White |first=R. W. |date=1959 |title=Motivation reconsidered: The concept of competence |journal=Psychological Review |volume=66 |issue=5 |pages=297–333 |doi=10.1037/h0040934 |pmid=13844397 |s2cid=37385966}}</ref>

Although the ability to sense (e.g. [[computer vision|see]], hear, etc.) and the ability to act (e.g. [[robotics|move and manipulate objects]], change location to explore, etc.) can be desirable for some intelligent systems,<ref name=":13"/> these physical capabilities are not strictly required for an entity to qualify as AGI—particularly under the thesis that large language models (LLMs) may already be or become AGI. Even from a less optimistic perspective on LLMs, there is no firm requirement for an AGI to have a human-like form; being a silicon-based computational system is sufficient, provided it can process input (language) from the external world in place of human senses. This interpretation aligns with the understanding that AGI has never been proscribed a particular physical embodiment and thus does not demand a capacity for locomotion or traditional "eyes and ears".<ref name="White 1959 297–333"/>  It can be regarded as sufficient for an intelligent computer to ''interact with other systems'', to invoke or regulate them, to achieve specific goals, including altering a physical environment, as [[HAL 9000|HAL]] in  ''[[2001: A Space Odyssey]]'' was both programmed and tasked to.<ref>{{cite web |url=http://www.robothalloffame.org/inductees/03inductees/hal.html |title=HAL 9000 |website=Robot Hall of Fame |publisher=Robot Hall of Fame, Carnegie Science Center |access-date=July 28, 2013 |archive-url=https://web.archive.org/web/20130917134208/http://www.robothalloffame.org/inductees/03inductees/hal.html |archive-date=September 17, 2013 |url-status=live}}</ref>

===Tests for human-level AGI{{Anchor|Tests_for_confirming_human-level_AGI}}===
Several tests meant to confirm human-level AGI have been considered, including:<ref>{{Cite web |last=Muehlhauser |first=Luke |date=11 August 2013 |title=What is AGI? |url=http://intelligence.org/2013/08/11/what-is-agi/ |url-status=live |archive-url=https://web.archive.org/web/20140425115445/http://intelligence.org/2013/08/11/what-is-agi/ |archive-date=25 April 2014 |access-date=1 May 2014 |publisher=Machine Intelligence Research Institute}}</ref><ref>{{Cite web |date=13 July 2019 |title=What is Artificial General Intelligence (AGI)? {{!}} 4 Tests For Ensuring Artificial General Intelligence |url=https://www.talkyblog.com/artificial_general_intelligence_agi/ |url-status=live |archive-url=https://web.archive.org/web/20190717071152/https://www.talkyblog.com/artificial_general_intelligence_agi/ |archive-date=17 July 2019 |access-date=17 July 2019 |website=Talky Blog |language=en-US}}</ref>

;[[Turing test|The Turing Test]] ([[Alan Turing|''Turing'']])  
:[[File:Weakness of Turing test 1.svg|thumb|The [[Turing test]] can provide some evidence of intelligence, but it penalizes non-human intelligent behavior and may incentivize [[artificial stupidity]].<ref>{{Cite magazine |last=Batson |first=Joshua |title=Forget the Turing Test: Here's How We Could Actually Measure AI |url=https://www.wired.com/2014/06/beyond-the-turing-test/ |access-date=2025-03-22 |magazine=Wired |language=en-US |issn=1059-1028}}</ref>]]Proposed by Alan Turing in his 1950 paper "Computing Machinery and Intelligence", this test involves a human judge engaging in natural language conversations with both a human and a machine designed to generate human-like responses. The machine passes the test if it can convince the judge it is human a significant fraction of the time. Turing proposed this as a practical measure of machine intelligence, focusing on the ability to produce human-like responses rather than on the internal workings of the machine.{{Sfn|Turing|1950}}

: Turing described the test as follows:
{{Blockquote|text=The idea of the test is that the machine has to try and pretend to be a man, by answering questions put to it, and it will only pass if the pretence is reasonably convincing. A considerable portion of a jury, who should not be expert about machines, must be taken in by the pretence.<ref name="Turing1952">{{Cite book |last=Turing |first=Alan |title=Can Automatic Calculating Machines Be Said To Think? |publisher=Oxford University Press |date=1952 |isbn=978-0-1982-5079-1 |editor-last=B. Jack Copeland |editor-link=Jack Copeland |publication-place=Oxford |pages=487–506}}</ref>}}

: In 2014, a chatbot named [[Eugene Goostman]], designed to imitate a 13-year-old Ukrainian boy, reportedly passed a Turing Test event by convincing 33% of judges that it was human. However, this claim was met with significant skepticism from the AI research community, who questioned the test's implementation and its relevance to AGI.<ref>{{Cite news |date=2014-06-09 |title=Eugene Goostman is a real boy – the Turing Test says so |url=https://www.theguardian.com/technology/shortcuts/2014/jun/09/eugene-goostman-turing-test-computer-program |access-date=2024-03-03 |work=The Guardian |language=en-GB |issn=0261-3077}}</ref><ref>{{Cite web |date=2014-06-09 |title=Scientists dispute whether computer 'Eugene Goostman' passed Turing test |url=https://www.bbc.com/news/technology-27762088 |access-date=2024-03-03 |website=BBC News}}</ref>

: In 2023, it was claimed that "AI is closer to ever" to passing the Turing test, though the article's authors reinforced that ''imitation'' (as "[[large language model]]s" ever closer to passing the test are built upon) is not synonymous with "intelligence".  Further, as AI intelligence and human intelligence may differ, "passing the Turing test is good evidence a system is intelligent, failing it is not good evidence a system is not intelligent."<ref>{{Cite web |last1=Kirk-Giannini |first1=Cameron Domenico |last2=Goldstein |first2=Simon |date=2023-10-16 |title=AI is closer than ever to passing the Turing test for 'intelligence'. What happens when it does? |url=https://theconversation.com/ai-is-closer-than-ever-to-passing-the-turing-test-for-intelligence-what-happens-when-it-does-214721 |access-date=2024-09-22 |website=The Conversation |language=en-US}}</ref>

: A 2024 study suggested that [[GPT-4]] was identified as human 54% of the time in a randomized, controlled version of the Turing Test—surpassing older chatbots like ELIZA while still falling behind actual humans (67%).<ref>{{Cite arXiv |last1=Jones |first1=Cameron R. |last2=Bergen |first2=Benjamin K. |title=People cannot distinguish GPT-4 from a human in a Turing test |eprint=2405.08007 |class=cs.HC  |date=9 May 2024 }}</ref>

: A 2025 pre‑registered, three‑party Turing‑test study by Cameron R. Jones and Benjamin K. Bergen showed that [[GPT-4.5]] was judged to be the human in 73% of five‑minute text conversations—surpassing the 67% humanness rate of real confederates and meeting the researchers’ criterion for having passed the test.<ref>{{cite arXiv |title=Large Language Models Pass the Turing Test |eprint=2503.23674 |date=2025-03-31 |last1=Jones |first1=Cameron R. |last2=Bergen |first2=Benjamin K. |class=cs.CL }}</ref><ref>{{cite news |title=AI model passes Turing Test better than a human |url=https://www.independent.co.uk/tech/ai-turing-test-chatgpt-openai-agi-b2728930.html |work=The Independent |date=2025-04-09 |access-date=2025-04-18}}</ref>

;The Robot College Student Test ([[Ben Goertzel|''Goertzel'']])  
: A machine enrolls in a university, taking and passing the same classes that humans would, and obtaining a degree. LLMs can now pass university degree-level exams without even attending the classes.<ref>{{Cite web |last=Varanasi |first=Lakshmi |date=21 March 2023 |title=AI models like ChatGPT and GPT-4 are acing everything from the bar exam to AP Biology. Here's a list of difficult exams both AI versions have passed. |url=https://www.businessinsider.com/list-here-are-the-exams-chatgpt-has-passed-so-far-2023-1 |access-date=30 May 2023 |website=[[Business Insider]]}}</ref>

;The Employment Test ([[Nils John Nilsson|''Nilsson'']])  
: A machine performs an economically important job at least as well as humans in the same job. AIs are now replacing humans in many roles as varied as fast food and marketing.<ref>{{Cite web |last=Naysmith |first=Caleb |date=7 February 2023 |title=6 Jobs Artificial Intelligence Is Already Replacing and How Investors Can Capitalize on It |url=https://www.yahoo.com/now/6-jobs-artificial-intelligence-already-150339825.html |access-date=30 May 2023}}</ref>

;The Ikea test ([[Gary Marcus|''Marcus'']])  
: Also known as the Flat Pack Furniture Test. An AI views the parts and instructions of an Ikea flat-pack product, then controls a robot to assemble the furniture correctly.<ref>{{Cite web |last=Turk |first=Victoria |date=2015-01-28 |title=The Plan to Replace the Turing Test with a 'Turing Olympics' |url=https://www.vice.com/en/article/the-plan-to-replace-the-turing-test-with-a-turing-olympics/ |access-date=2024-03-03 |website=Vice |language=en}}</ref>

;The Coffee Test ([[Steve Wozniak|''Wozniak'']])  
: A machine is required to enter an average American home and figure out how to make coffee: find the coffee machine, find the coffee, add water, find a mug, and brew the coffee by pushing the proper buttons.<ref>{{Cite web |last=Gopani |first=Avi |date=2022-05-25 |title=Turing Test is unreliable. The Winograd Schema is obsolete. Coffee is the answer. |url=https://analyticsindiamag.com/turing-test-is-unreliable-the-winograd-schema-is-obsolete-coffee-is-the-answer/ |access-date=2024-03-03 |website=Analytics India Magazine |language=en-US}}</ref> This has not yet been completed.

;The Modern Turing Test (''[[Mustafa Suleyman|Suleyman]]'')  
: An AI model is given $100,000 and has to obtain $1&nbsp;million.<ref>{{Cite web |last=Bhaimiya |first=Sawdah |date=June 20, 2023 |title=DeepMind's co-founder suggested testing an AI chatbot's ability to turn $100,000 into $1 million to measure human-like intelligence |url=https://www.businessinsider.com/deepmind-co-founder-suggests-new-turing-test-ai-chatbots-report-2023-6 |access-date=2024-03-03 |website=Business Insider |language=en-US}}</ref><ref>{{Cite web |last=Suleyman |first=Mustafa |date=July 14, 2023 |title=Mustafa Suleyman: My new Turing test would see if AI can make $1 million |url=https://www.technologyreview.com/2023/07/14/1076296/mustafa-suleyman-my-new-turing-test-would-see-if-ai-can-make-1-million/ |access-date=2024-03-03 |website=MIT Technology Review |language=en}}</ref>

===AI-complete problems===
{{Main|AI-complete}}

A problem is informally called "AI-complete" or "AI-hard" if it is believed that in order to solve it, one would need to implement AGI, because the solution is beyond the capabilities of a purpose-specific algorithm.<ref name="Shapiro92">{{Cite book |last=Shapiro |first=Stuart C. |title=Encyclopedia of Artificial Intelligence |publisher=John Wiley |date=1992 |editor-last=Stuart C. Shapiro |edition=Second |location=New York |pages=54–57 |chapter=Artificial Intelligence |chapter-url=http://www.cse.buffalo.edu/~shapiro/Papers/ai.pdf |archive-url=https://web.archive.org/web/20160201014644/http://www.cse.buffalo.edu/~shapiro/Papers/ai.pdf |archive-date=1 February 2016 |url-status=live}} (Section 4 is on "AI-Complete Tasks".)</ref>

There are many problems that have been conjectured to require general intelligence to solve as well as humans. Examples include [[computer vision]], [[natural-language understanding|natural language understanding]], and dealing with unexpected circumstances while solving any real-world problem.<ref>{{Cite journal |last=Yampolskiy |first=Roman V. |date=2012 |title=Turing Test as a Defining Feature of AI-Completeness |url=http://cecs.louisville.edu/ry/TuringTestasaDefiningFeature04270003.pdf |url-status=live |journal=Artificial Intelligence, Evolutionary Computation and Metaheuristics  |pages=3–17 |archive-url=https://web.archive.org/web/20130522094547/http://cecs.louisville.edu/ry/TuringTestasaDefiningFeature04270003.pdf |archive-date=22 May 2013 |editor=Xin-She Yang}}</ref> Even a specific task like [[machine translation|translation]] requires a machine to read and write in both languages, follow the author's argument (reason), understand the context (knowledge), and faithfully reproduce the author's original intent ([[social intelligence]]). All of these problems need to be solved simultaneously in order to reach human-level machine performance.

However, many of these tasks can now be performed by modern large language models. According to [[Stanford University]]'s 2024 AI index, AI has reached human-level performance on many [[Benchmarks for artificial intelligence|benchmarks]] for reading comprehension and visual reasoning.<ref>{{Cite web |date=2024-04-15 |title=AI Index: State of AI in 13 Charts |url=https://hai.stanford.edu/news/ai-index-state-ai-13-charts |access-date=2024-05-27 |website=Stanford University Human-Centered Artificial Intelligence |language=en}}</ref>