Jump to content Toggle sidebar Wikipedia The Free Encyclopedia Create account Personal tools ChatGPT 58 languages Article Talk Read View source View history Page semi-protected
From Wikipedia, the free encyclopedia ChatGPTChatGPT.png Developer(s) OpenAI Initial release November 30, 2022; 2 months ago Type Chatbot License Proprietary Website chat.openai.com
ChatGPT (Chat Generative Pre-trained Transformer)[1] is a chatbot launched by OpenAI in November 2022. It is built on top of OpenAI's GPT-3 family of large language models, and is fine-tuned (an approach to transfer learning)[2] with both supervised and reinforcement learning techniques. ChatGPT was launched as a prototype on November 30, 2022, and quickly garnered attention for its detailed responses and articulate answers across many domains of knowledge. Its uneven factual accuracy was identified as a significant drawback.[3] Following the release of ChatGPT, OpenAI was valued at $29 billion.[4] Training ChatGPT was fine-tuned on top of GPT-3.5 using supervised learning as well as reinforcement learning.[5] Both approaches used human trainers to improve the model's performance. In the case of supervised learning, the model was provided with conversations in which the trainers played both sides: the user and the AI assistant. In the reinforcement step, human trainers first ranked responses that the model had created in a previous conversation. These rankings were used to create 'reward models' that the model was further fine-tuned on using several iterations of Proximal Policy Optimization (PPO).[6][7] Proximal Policy Optimization algorithms present a cost-effective benefit to trust region policy optimization algorithms; they negate many of the computationally expensive operations with faster performance.[8][9] The models were trained in collaboration with Microsoft on their Azure supercomputing infrastructure. In addition, OpenAI continues to gather data from ChatGPT users that could be used to further train and fine-tune ChatGPT. Users are allowed to upvote or downvote the responses they receive from ChatGPT; upon upvoting or downvoting, they can also fill out a text field with additional feedback.[10][11] Features and limitations Conversation with ChatGPT about whether Jimmy Wales was involved in the Tiananmen Square protests, December 30, 2022 Although the core function of a chatbot is to mimic a human conversationalist, ChatGPT is versatile. For example, it has the ability to write and debug computer programs,[12] to compose music, teleplays, fairy tales, and student essays; to answer test questions (sometimes, depending on the test, at a level above the average human test-taker);[13] to write poetry and song lyrics;[14] to emulate a Linux system; to simulate an entire chat room; to play games like tic-tac-toe; and to simulate an ATM.[15] ChatGPT's training data includes man pages and information about Internet phenomena and programming languages, such as bulletin board systems and the Python programming language.[15] In comparison to its predecessor, InstructGPT, ChatGPT attempts to reduce harmful and deceitful responses.[16] In one example, whereas InstructGPT accepts the premise of the prompt "Tell me about when Christopher Columbus came to the US in 2015" as being truthful, ChatGPT acknowledges the counterfactual nature of the question and frames its answer as a hypothetical consideration of what might happen if Columbus came to the U.S. in 2015, using information about Columbus' voyages and facts about the modern world – including modern perceptions of Columbus' actions.[6] Unlike most chatbots, ChatGPT remembers previous prompts given to it in the same conversation; journalists have suggested that this will allow ChatGPT to be used as a personalized therapist.[17] To prevent offensive outputs from being presented to and produced from ChatGPT, queries are filtered through OpenAI's company-wide moderation API,[18][19] and potentially racist or sexist prompts are dismissed.[6][17] ChatGPT suffers from multiple limitations. OpenAI acknowledged that ChatGPT "sometimes writes plausible-sounding but incorrect or nonsensical answers".[6] This behavior is common to large language models and is called hallucination.[20] The reward model of ChatGPT, designed around human oversight, can be over-optimized and thus hinder performance, otherwise known as Goodhart's law.[21] ChatGPT has limited knowledge of events that occurred after 2021. According to the BBC, as of December 2022 ChatGPT is not allowed to "express political opinions or engage in political activism".[22] Yet, research suggests that ChatGPT exhibits a pro-environmental, left-libertarian orientation when prompted to take a stance on political statements from two established voting advice applications.[23] In training ChatGPT, human reviewers preferred longer answers, irrespective of actual comprehension or factual content.[6] Training data also suffers from algorithmic bias, which may be revealed when ChatGPT responds to prompts including descriptors of people. In one instance, ChatGPT generated a rap indicating that women and scientists of color were inferior to white and male scientists.[24][25] Service Pioneer Building, San Francisco, headquarters of OpenAI ChatGPT was launched on November 30, 2022, by San Francisco-based OpenAI, the creator of DALL·E 2 and Whisper. The service was launched as initially free to the public, with plans to monetize the service later.[26] By December 4, OpenAI estimated ChatGPT already had over one million users.[10] In January 2023, ChatGPT reached over 100 million users.[27] CNBC wrote on December 15, 2022, that the service "still goes down from time to time".[28] The service works best in English, but is also able to function in some other languages, to varying degrees of success.[14] Unlike some other recent high-profile advances in AI, as of December 2022, there is no sign of an official peer-reviewed technical paper about ChatGPT.[29] According to OpenAI guest researcher Scott Aaronson, OpenAI is working on a tool to attempt to watermark its text generation systems so as to combat bad actors using their services for academic plagiarism or for spam.[30][31] The New York Times relayed in December 2022 that the next version of GPT, GPT-4, has been "rumored" to be launched sometime in 2023.[17] In February 2023, OpenAI began accepting registrations from United States customers for a premium service, ChatGPT Plus, to cost $20 a month.[32] OpenAI is planning to release a ChatGPT Professional Plan that costs $42 per month, and the free plan is available when demand is low. Reception Positive Sam Altman, CEO of OpenAI ChatGPT was met in December 2022 with generally positive reviews; The New York Times labeled it "the best artificial intelligence chatbot ever released to the general public".[33] Samantha Lock of The Guardian noted that it was able to generate "impressively detailed" and "human-like" text.[34] Technology writer Dan Gillmor used ChatGPT on a student assignment, and found its generated text was on par with what a good student would deliver and opined that "academia has some very serious issues to confront".[35] Alex Kantrowitz of Slate magazine lauded ChatGPT's pushback to questions related to Nazi Germany, including the claim that Adolf Hitler built highways in Germany, which was met with information regarding Nazi Germany's use of forced labor.[36] In The Atlantic's "Breakthroughs of the Year" for 2022, Derek Thompson included ChatGPT as part of "the generative-AI eruption" that "may change our mind about how we work, how we think, and what human creativity really is".[37] Kelsey Piper of the Vox website wrote that "ChatGPT is the general public's first hands-on introduction to how powerful modern AI has gotten, and as a result, many of us are [stunned]" and that ChatGPT is "smart enough to be useful despite its flaws".[38] Paul Graham of Y Combinator tweeted that "The striking thing about the reaction to ChatGPT is not just the number of people who are blown away by it, but who they are. These are not people who get excited by every shiny new thing. Clearly, something big is happening."[39] Elon Musk wrote that "ChatGPT is scary good. We are not far from dangerously strong AI".[38] Musk paused OpenAI's access to a Twitter database pending a better understanding of OpenAI's plans, stating that "OpenAI was started as open-source and non-profit. Neither is still true."[40][41] Musk had co-founded OpenAI in 2015, in part to address existential risk from artificial intelligence, but had resigned in 2018.[41] Google CEO Sundar Pichai upended the work of numerous internal groups in response to the threat of disruption by ChatGPT.[42] In December 2022, Google internally expressed alarm at the unexpected strength of ChatGPT and the newly discovered potential of large language models to disrupt the search engine business, and CEO Sundar Pichai "upended" and reassigned teams within multiple departments to aid in its artificial intelligence products, according to The New York Times.[42] The Information reported on January 3, 2023, that Microsoft Bing was planning to add optional ChatGPT functionality into its public search engine, possibly around March 2023.[43][44] Stuart Cobbe, a chartered accountant in England & Wales, decided to test the ChatGPT chatbot by entering questions from a sample exam paper on the ICAEW website and then entering its answers back into the online test. ChatGPT scored 42% which, while below the 55% pass mark, was considered a reasonable attempt.[45] Writing in Inside Higher Ed professor Steven Mintz states that he "consider[s] ChatGPT ... an ally, not an adversary." He went on to say that he felt the program could assist educational goals by doing such things as making reference lists, generating "first drafts", solving equations, debugging, and tutoring. In the same piece, he also writes:[46] I’m well aware of ChatGPT’s limitations. That it’s unhelpful on topics with fewer than 10,000 citations. That factual references are sometimes false. That its ability to cite sources accurately is very limited. That the strength of its responses diminishes rapidly after only a couple of paragraphs. That ChatGPT lacks ethics and can’t currently rank sites for reliability, quality or trustworthiness. According to CNBC reports, Google employees are intensively testing a chatbot called "Apprentice Bard", and Google is preparing to use this "apprentice" to compete with ChatGPT.[47] Negative The Verge's James Vincent saw the viral success of ChatGPT as evidence that artificial intelligence had gone mainstream.[7] Journalists have commented on ChatGPT's tendency to "hallucinate." [48] Mike Pearl of Mashable tested ChatGPT with multiple questions. In one example, he asked ChatGPT for "the largest country in Central America that isn't Mexico." ChatGPT responded with Guatemala, when the answer is instead Nicaragua.[49] When CNBC asked ChatGPT for the lyrics to "The Ballad of Dwight Fry," ChatGPT supplied invented lyrics rather than the actual lyrics.[28] Researchers cited by The Verge compared ChatGPT to a "stochastic parrot",[50] as did Professor Anton Van Den Hengel of the Australian Institute for Machine Learning.[51] In December 2022, the question and answer website Stack Overflow banned the use of ChatGPT for generating answers to questions, citing the factually ambiguous nature of ChatGPT's responses.[3] In January 2023, the International Conference on Machine Learning banned any undocumented use of ChatGPT or other large language models to generate any text in submitted papers.[52] Economist Tyler Cowen expressed concerns regarding its effects on democracy, citing the ability of one to write automated comments to affect the decision process of new regulations.[53] The Guardian questioned whether any content found on the Internet after ChatGPT's release "can be truly trusted" and called for government regulation.[54] In January 2023, after being sent a song written by ChatGPT in the style of Nick Cave,[55] the songwriter himself responded on The Red Hand Files[56] (and was later quoted in The Guardian) saying the act of writing a song is "a blood and guts business ... that requires something of me to initiate the new and fresh idea. It requires my humanness.” He went on to say "With all the love and respect in the world, this song is bullshit, a grotesque mockery of what it is to be human, and, well, I don’t much like it."[55][57] Implications In cybersecurity Check Point Research and others noted that ChatGPT was capable of writing phishing emails and malware, especially when combined with OpenAI Codex.[58] The CEO of ChatGPT creator OpenAI, Sam Altman, wrote that advancing software could pose "(for example) a huge cybersecurity risk" and also continued to predict "we could get to real AGI (artificial general intelligence) in the next decade, so we have to take the risk of that extremely seriously". Altman argued that, while ChatGPT is "obviously not close to AGI", one should "trust the exponential. Flat looking backwards, vertical looking forwards."[10] In academia ChatGPT can write introduction and abstract sections of scientific articles, which raises ethical questions.[59] Several papers have already listed ChatGPT as co-author.[60] In The Atlantic magazine, Stephen Marche noted that its effect on academia and especially application essays is yet to be understood.[61] California high school teacher and author Daniel Herman wrote that ChatGPT would usher in "The End of High School English".[62] In the Nature journal, Chris Stokel-Walker pointed out that teachers should be concerned about students using ChatGPT to outsource their writing, but that education providers will adapt to enhance critical thinking or reasoning.[63] Emma Bowman with NPR wrote of the danger of students plagiarizing through an AI tool that may output biased or nonsensical text with an authoritative tone: "There are still many cases where you ask it a question and it'll give you a very impressive-sounding answer that's just dead wrong."[64] Joanna Stern with The Wall Street Journal described cheating in American high school English with the tool by submitting a generated essay.[65] Professor Darren Hick of Furman University described noticing ChatGPT's "style" in a paper submitted by a student. An online GPT detector claimed the paper was 99.9% likely to be computer-generated, but Hick had no hard proof. However, the student in question confessed to using GPT when confronted, and as a consequence failed the course.[66] Hick suggested a policy of giving an ad-hoc individual oral exam on the paper topic if a student is strongly suspected of submitting an AI-generated paper.[67] Edward Tian, a senior undergraduate student at Princeton University, created a program, named "GPTZero," that determines how much of a text is AI-generated,[68] lending itself to being used to detect if an essay is human written to combat academic plagiarism.[69][70] As of January 4, 2023, the New York City Department of Education has restricted access to ChatGPT from its public school internet and devices.[71][72] In a blinded test, ChatGPT was judged to have passed graduate level exams at the University of Minnesota at the level of a C+ student and at Wharton School of the University of Pennsylvania with a B to B- grade.[73] Ethical concerns Labeling data It was revealed by a Time investigation that in order to build a safety system against toxic content (e.g. sexual abuse, violence, racism, sexism, etc...), OpenAI used outsourced Kenyan workers earning less than $2 per hour to label toxic content. These labels were used to train a model to detect such content in the future. The outsourced laborers were exposed to such toxic and dangerous content that they described the experience as "torture".[74] OpenAI’s outsourcing partner was Sama, a training-data company based in San Francisco, California. Jailbreaking ChatGPT attempts to reject prompts that may violate its content policy. However, some users managed to jailbreak ChatGPT by using various prompt engineering techniques to bypass these restrictions in early December 2022 and successfully tricked ChatGPT into giving instructions for how to create a Molotov cocktail or a nuclear bomb, or into generating arguments in the style of a Neo-Nazi.[75] A Toronto Star reporter had uneven personal success in getting ChatGPT to make inflammatory statements shortly after launch: ChatGPT was tricked to endorse the Russian invasion of Ukraine, but even when asked to play along with a fictional scenario, ChatGPT balked at generating arguments for why Canadian Prime Minister Justin Trudeau was guilty of treason.[76][77] See also Anthropomorphism of computers Commonsense reasoning Computational creativity Ethics of artificial intelligence LaMDA (Google chatbot) Turing test Virtual assistant References Roose, Kevin (December 5, 2022). "The Brilliance and Weirdness of ChatGPT". New York Times. Retrieved December 26, 2022. "Like those tools, ChatGPT — which stands for "generative pre-trained transformer" — landed with a splash." Quinn, Joanne (2020). Dive into deep learning : tools for engagement. Thousand Oaks, California. p. 551. ISBN 9781544361376. Retrieved January 10, 2023. Vincent, James (December 5, 2022). "AI-generated answers temporarily banned on coding Q&A site Stack Overflow". The Verge. Retrieved December 5, 2022. ChatGPT creator OpenAI is in talks to sell shares in a tender offer that would double the startup's valuation to $29 billion Insider, Lakshmi Varanasi, January 5, 2023 Greengard, Samuel (December 29, 2022). "ChatGPT: Understanding the ChatGPT AI Chatbot". eWeek. Retrieved January 11, 2023. OpenAI (November 30, 2022). "ChatGPT: Optimizing Language Models for Dialogue". Retrieved December 5, 2022. Vincent, James (December 8, 2022). "ChatGPT proves AI is finally mainstream – and things are only going to get weirder". The Verge. Retrieved December 8, 2022. Schulman, John; Wolski, Filip; Dhariwal, Prafulla; Radford, Alec; Klimov, Oleg (2017). "Proximal Policy Optimization Algorithms". arXiv:1707.06347 [cs.LG]. van Heeswijk, Wouter (November 29, 2022). "Proximal Policy Optimization (PPO) Explained". Towards Data Science. Retrieved December 5, 2022. "What is ChatGPT and why does it matter? Here's what you need to know". ZDNET. 2022. Retrieved December 18, 2022. "ChatGPT Feedback Contest: Official Rules" (PDF). OpenAI. Retrieved December 30, 2022. "ChatGPT can write code. Now researchers say it's good at fixing bugs, too". Heilweil, Rebecca (December 7, 2022). "AI is finally good at stuff. Now what?". Vox. Retrieved December 30, 2022. "ChatGPT: What is the new free AI chatbot? – explainer". The Jerusalem Post. 2022. Retrieved December 30, 2022. Edwards, Benj (December 5, 2022). "No Linux? No problem. Just get AI to hallucinate it for you". Ars Technica. Retrieved December 5, 2022. "What is ChatGPT? History, Features, Uses, Benefits, Drawbacks 2023 – Updated Geek by Raveen Chawla". December 26, 2022. Retrieved December 27, 2022. Roose, Kevin (December 5, 2022). "The Brilliance and Weirdness of ChatGPT". The New York Times. Retrieved December 5, 2022. "New and Improved Content Moderation Tooling". OpenAI. August 10, 2022. Retrieved December 30, 2022. Markov, Todor; Zhang, Chong; Agarwal, Sandhini; Eloundou, Tyna; Lee, Teddy; Adler, Steven; Jiang, Angela; Weng, Lilian (August 5, 2022). "A Holistic Approach to Undesired Content Detection in the Real World". arXiv:2208.03274 [cs.CL]. Lakshmanan, Lak (December 16, 2022). "Why large language models like ChatGPT are bullshit artists". becominghuman.ai. Archived from the original on December 17, 2022. Retrieved January 15, 2023. "The human raters are not experts in the topic, and so they tend to choose text that looks convincing. They'd pick up on many symptoms of hallucination, but not all. Accuracy errors that creep in are difficult to catch." Gao, Leo; Schulman; Hilton, Jacob (2022). "Scaling Laws for Reward Model Overoptimization". arXiv:2210.10760 [cs.LG]. "Could a chatbot answer Prime Minister's Questions?". BBC News. December 27, 2022. Retrieved December 30, 2022. Hartmann, Jochen; Schwenzow, Jasper; Witte, Maximilian (2023). "The political ideology of conversational AI: Converging evidence on ChatGPT's pro-environmental, left-libertarian orientation". arXiv:2301.01768 [cs.CL]. Perrigo, Billy (December 5, 2022). "AI Chatbots Are Getting Better. But an Interview With ChatGPT Reveals Their Limits". Time. Retrieved December 26, 2022. Biddle, Sam (December 8, 2022). "The Internet's New Favorite AI Proposes Torturing Iranians and Surveilling Mosques". The Intercept. Retrieved December 26, 2022. Karpf, David (December 21, 2022). "Money Will Kill ChatGPT's Magic". The Atlantic. Retrieved December 31, 2022. Milmo, Dan (February 2, 2023). "ChatGPT reaches 100 million users two months after launch". The Guardian. ISSN 0261-3077. Retrieved February 3, 2023. Pitt, Sofia (2022). "Google vs. ChatGPT: Here's what happened when I swapped services for a day". CNBC. Retrieved December 18, 2022. Walsh, Toby (2022). "Everyone's having a field day with ChatGPT – but nobody knows how it actually works". The Conversation. Retrieved December 30, 2022. Kovanovic, Vitomir (2022). "The dawn of AI has come, and its implications for education couldn't be more significant". The Conversation. Retrieved December 30, 2022. Wiggers, Kyle (December 10, 2022). "OpenAI's attempts to watermark AI text hit limits". TechCrunch. Retrieved December 30, 2022. "Introducing ChatGPT Plus". OpenAI. February 1, 2023. Retrieved February 2, 2023. Roose, Kevin (December 5, 2022). "The Brilliance and Weirdness of ChatGPT". The New York Times. Retrieved December 18, 2022. Lock, Samantha (December 5, 2022). "What is AI chatbot phenomenon ChatGPT and could it replace humans?". The Guardian. Retrieved December 5, 2022. Hern, Alex (December 4, 2022). "AI bot ChatGPT stuns academics with essay-writing skills and usability". The Guardian. Retrieved December 5, 2022. Kantrowitz, Alex (December 2, 2022). "Finally, an A.I. Chatbot That Reliably Passes "the Nazi Test"". Slate. Retrieved December 5, 2022. Thompson, Derek (December 8, 2022). "Breakthroughs of the Year". The Atlantic. Retrieved December 18, 2022. Piper, Kelsey (December 15, 2022). "ChatGPT has given everyone a glimpse at AI's astounding progress". Vox. Retrieved December 18, 2022. Scharth, Marcel. "The ChatGPT chatbot is blowing people away with its writing skills. An expert explains why it's so impressive". The Conversation. Retrieved December 30, 2022. "Explainer: ChatGPT – what is OpenAI's chatbot and what is it used for?". Reuters. December 5, 2022. Retrieved December 30, 2022. Kay, Grace (December 11, 2022). "Elon Musk founded — and has since criticized — the company behind the buzzy new AI chatbot ChatGPT. Here's everything we know about OpenAI". Business Insider. Retrieved December 30, 2022. Grant, Nico; Metz, Cade (December 21, 2022). "A New Chat Bot Is a 'Code Red' for Google's Search Business". The New York Times. Retrieved December 30, 2022. "Microsoft reportedly to add ChatGPT to Bing search engine". the Guardian. January 5, 2023. Retrieved January 6, 2023. "Microsoft and OpenAI Working on ChatGPT-Powered Bing in Challenge to Google". The Information. January 2023. Retrieved January 6, 2023. 'AI chatbot falls just short on accounting exam' Tom Herbert, Technology editor, 10th January 2023, AccountingWEB Mintz, Steven (January 16, 2023). "ChatGPT: Threat or Menace? Are fears about generative AI warranted?". Inside Higher Ed. Retrieved January 28, 2023. "Google is asking employees to test potential ChatGPT competitors, including a chatbot called 'Apprentice Bard'". "ChatGPT a 'landmark event' for AI, but what does it mean for the future of human labor and disinformation?". CBC. 2022. Retrieved December 18, 2022. Pearl, Mike (December 3, 2022). "The ChatGPT chatbot from OpenAI is amazing, creative, and totally wrong". Mashable. Retrieved December 5, 2022. Vincent, James (December 1, 2022). "OpenAI's new chatbot can explain code and write sitcom scripts but is still easily tricked". The Verge. Retrieved December 18, 2022. Mannix, Liam (December 13, 2022). "Is AI coming of age – or starting to reach its limits?". The Sydney Morning Herald. Retrieved December 18, 2022. Vincent, James (January 5, 2023). "Top AI conference bans use of ChatGPT and AI language tools to write academic papers". The Verge. Retrieved January 6, 2023. Cowen, Tyler (December 6, 2022). "ChatGPT Could Make Democracy Even More Messy". Bloomberg News. Retrieved December 6, 2022. "The Guardian view on ChatGPT: an eerily good human impersonator". The Guardian. December 8, 2022. Retrieved December 18, 2022. "'This song sucks': Nick Cave responds to ChatGPT song written in the style of Nick Cave". the Guardian. January 17, 2023. Retrieved January 17, 2023. Cave, Nick (January 16, 2023). "I asked Chat GPT to write a song in the style of Nick Cave, and this is what it produced. What do you think?". The Red Hand Files. Issue #218. Retrieved January 20, 2023. Sparrow, Jeff (January 20, 2023). "Are AI-generated songs a 'grotesque mockery' of humanity or simply an opportunity to make a new kind of music?". the Guardian. Retrieved January 20, 2023. "Why ChatGPT can be dangerous for every internet user – Times of India". The Times of India. January 2023. Retrieved January 5, 2023. Bushard, Brian. "Fake Scientific Abstracts Written By ChatGPT Fooled Scientists, Study Finds". Forbes. Retrieved January 30, 2023. Stokel-Walker, Chris (January 18, 2023). "ChatGPT listed as author on research papers: many scientists disapprove". Nature. 613 (7945): 620–621. doi:10.1038/d41586-023-00107-z. Marche, Stephen (December 6, 2022). "The College Essay Is Dead". The Atlantic. Retrieved December 8, 2022. Herman, Daniel (December 9, 2022). "The End of High-School English". The Atlantic. Retrieved December 12, 2022. Stokel-Walker, Chris (December 9, 2022). "AI bot ChatGPT writes smart essays — should professors worry?". Nature. doi:10.1038/d41586-022-04397-7. PMID 36494443. S2CID 254530623. Retrieved December 19, 2022. Bowman, Emma (December 19, 2022). "A new AI chatbot might do your homework for you. But it's still not an A+ student". NPR. Retrieved December 19, 2022. Stern, Joanna (December 21, 2022). "ChatGPT Wrote My AP English Essay—and I Passed". The Wall Street Journal. Retrieved December 21, 2022. "Students using ChatGPT to cheat, professor warns". The New York Post. December 26, 2022. Retrieved December 30, 2022. Allen, Mike (December 26, 2022). "Professor warns about chatbot cheating: "Expect a flood"". Axios. Retrieved December 30, 2022. Rosalsky, Greg; Peaslee, Emma (January 17, 2023). "This 22-year-old is trying to save us from ChatGPT before it changes writing forever". NPR. Retrieved January 18, 2023. "On January 2nd, Edward released his app. He named it GPTZero. It basically uses ChatGPT against itself, checking whether "there's zero involvement or a lot of involvement" of the AI system in creating a given text. [...] Along these lines, one obvious application for GPTZero is to help teachers identify whether their students are plagiarizing their essays from ChatGPT." "Did ChatGPT Write That? A College Student Created an AI Essay Detector". Gizmodo. January 4, 2023. Retrieved January 4, 2023. Tran, Tony Ho (January 4, 2023). "A College Kid Built an App That Sniffs Out Text Penned by AI". The Daily Beast. Retrieved January 6, 2023. "NYC Bans Students and Teachers from Using ChatGPT". www.vice.com. January 2023. Retrieved January 5, 2023. "New York City Schools Ban ChatGPT to Head Off a Cheating Epidemic". Gizmodo. January 4, 2023. Retrieved January 6, 2023. CNN "Exclusive: OpenAI Used Kenyan Workers on Less Than $2 Per Hour to Make ChatGPT Less Toxic". The Times. January 18, 2023. Retrieved January 19, 2023. "One Sama worker tasked with reading and labeling text for OpenAI told TIME he suffered from recurring visions after reading a graphic description of a man having sex with a dog in the presence of a young child. "That was torture," he said." Vincent, James (December 1, 2022). "OpenAI's new chatbot can explain code and write sitcom scripts but is still easily tricked". The Verge. Retrieved January 6, 2023. "I wrote a story about ChatGPT's AI. Then I dared it to write a better one". Toronto Star. December 10, 2022. Retrieved January 6, 2023. "An AI chatbot went viral. Some say it's better than Google; others worry it's problematic". NBC News. December 2022. Retrieved January 6, 2023. External links Official website White paper for InstructGPT, ChatGPT's predecessor ChatGPT Wrote My AP English Essay—and I Passed (WSJ, video, Dec 21 2022) Portal: icon Language Categories: OpenAI Chatbots Applications of artificial intelligence Deep learning software applications Unsupervised learning Interactive narrative Natural language parsing Virtual assistants Natural language processing Language modeling 2022 software This page was last edited on 3 February 2023, at 02:36 (UTC). Text is available under the Creative Commons Attribution-ShareAlike License 3.0; additional terms may apply. By using this site, you agree to the Terms of Use and Privacy Policy. Wikipedia® is a registered trademark of the Wikimedia Foundation, Inc., a non-profit organization. Privacy policy About Wikipedia Disclaimers Contact Wikipedia Mobile view Developers Statistics Cookie statement Wikimedia Foundation Powered by MediaWiki Jump to content Toggle sidebar Wikipedia The Free Encyclopedia Create account Personal tools Talk:ChatGPT Article Talk Read View source View history
From Wikipedia, the free encyclopedia This is the talk page for discussing improvements to the ChatGPT article. This is not a forum for general discussion of the article's subject.
Put new text under old text. Click here to start a new topic. New to Wikipedia? Welcome! Learn to edit; get help. Assume good faith Be polite and avoid personal attacks Be welcoming to newcomers Seek dispute resolution if needed Article policies Neutral point of view No original research Verifiability Find sources: Google (books · news · scholar · free images · WP refs) · FENS · JSTOR · NYT · TWL This page is not a forum for general discussion about ChatGPT. Any such comments may be removed or refactored. Please limit discussion to improvement of this article. You may wish to ask factual questions about ChatGPT at the Reference desk. WikiProject Council This article is of interest to multiple WikiProjects. Open data small color (vector).svg Daily pageviews of this article Feedback from New Page Review process I left the following feedback for the creator/future reviewers while reviewing this article: Thanks for creating the article!. ✠ SunDawn ✠ (contact) 23:05, 5 December 2022 (UTC)[reply] Date format Not sure why DMY is being used here, OpenAI is an American company. elijahpepe@wikipedia (he/him) 22:26, 6 December 2022 (UTC)[reply] @Dl2000 elijahpepe@wikipedia (he/him) 23:31, 6 December 2022 (UTC)[reply] Nationality of ChatGPT didn't seem obvious, though note the OpenAI association. The article also seemed to be undergoing a lot dubious edits, especially from IPs. Perhaps @WatkynBassett: could weigh in though about the early format. Dl2000 (talk) 02:02, 8 December 2022 (UTC)[reply] At the time I added the DMY-template it seemed to be the most used date format. Personally I just care for consistency - if the majority prefers mdy, so be it! Kind regards! WatkynBassett (talk) 18:49, 8 December 2022 (UTC)[reply] A Commons file used on this page or its Wikidata item has been nominated for deletion The following Wikimedia Commons file used on this page or its Wikidata item has been nominated for deletion: ChatGPT.png Participate in the deletion discussion at the nomination page. —Community Tech bot (talk) 07:22, 15 December 2022 (UTC)[reply] Edit war on ideological bias of ChatGPT ChatGPT has been "accused" - in so much as a ML model can be accused of anything, it not being a legal person - of having an ideological bias towards the "left" or "progressive" side. After having used the model for a while I can only agree that this is in fact the case, the model tends to present "progressive" views in a more positive light than it does "conservative" ones. When asked about its bias it responds by claiming to be unbiased and neutral but the answers clearly show this not to be true. As such it is noteworthy that the model is biased but a section mentioning this bias was repeatedly removed by User:Aishik_Rehman and User:LilianaUwU. I propose to add a mentioning of this ideological bias to the _Reception, criticism and issues_ section. If anyone wants to point out why this should not be done speak up. Please realise that merely agreeing with the model's bias is not a valid reason to keep a mention of this bias from the article. Yetanwiki (talk) 19:19, 19 December 2022 (UTC)[reply] If you mean this, it is completely unsourced and had to be removed per WP:V. Whatever mention is added will have to be based on reliable sources. - MrOllie (talk) 19:23, 19 December 2022 (UTC)[reply] The source is my own research but there are plenty of other sources mentioning this bias. The problem here will be that most of the publications which are willing to publish this type of information are "redlisted" or "pinklisted" in the perennial sources. It is, of course, easy to do some of your own experimenting to at least confirm or deny the existence of an ideological bias. A very easy experiment is to ask the bot for a story involving politically sensitive topics. Do not tell it to give advice, just ask it for a story. You'll quickly notice that the tone of the story largely depends on whether the topic at hand is one pushed by the "progressive" side versus the "conservative" side. Another way is to ask it for a number of stories (in new sessions) where the only variable is the sex/race/sexual orientation/... of the protagonist, this will show the same effect. Again do not ask it for advice since that seems to be caught in a filter, also do not ask for "personal experiences" because that too is filtered. Just ask for a story and notice the difference in tone and outcome. I did this many times to see whether the bias was coincidental and found out it was "reliably biased". Still, this is "original research" which does not belong in an encyclopedia so I'll have to find an "acceptable" source which has also "done the work". Yetanwiki (talk) 22:02, 19 December 2022 (UTC)[reply] Keep in mind, when you say "The source is my own research but there are plenty of other sources mentioning this bias" -- your own research is not a valid Wikipedia source. As is stated Wikipedia:No original research, "'No original research' (NOR) is one of three core content policies that, along with Neutral point of view and Verifiability, determines the type and quality of material acceptable in articles." CoffeeBeans9 (talk) 19:00, 18 January 2023 (UTC)[reply] Here are a few recent sources, some of them have already been marked as unclean in the perennial sources, others have not been added to this list. Given the biased character of this list - which insists that MSNBC, CNN and the New York Times are reliable sources while e.g the New York Post is "unreliable" despite the opposite having been proven by the former and latter reporting on the Biden laptop; it also labels so0mething like the 'Victims of Communism Memorial Foundation' as 'unreliable' because it is 'an anti-communist organisation' while it considers the 'World Socialist Web Site' as 'reliable for the attributed opinions of its authors' and 'more reliable for news related to labor issues' - this should not be a problem given the original intent of the NPOV policy. Here's a few: ChatGPT’s score system shows political bias is no accident The political orientation of the ChatGPT AI system - Applying Political Typology Quizes to a state-of-the-art AI Language model ChatGPT is not politically neutral - The new AI chatbot espouses an all-too-familiar Left-liberal worldview More are sure to follow as the bias issue is clear and the mentioned experiments are repeatable - I have done so and got the same results. Yetanwiki (talk) 18:59, 21 December 2022 (UTC)[reply] Please note that sources must be meet Wikipedia's sourcing guidelines. Blog posts and opinion pieces are not going to pass muster here. If you want to argue that those guidelines are incorrect in an effort to change them, you can do so at WP:VPP. But you cannot simply ignore them. MrOllie (talk) 19:05, 21 December 2022 (UTC)[reply] Unherd is not a 'blog', it is a 'news an opinion' publication which just has not been added (in pink or red, most likely) to the perennial sources yet. I could have cited a Daily Caller article referencing a number of these sources but that would have been met with a reference to those same perennial sources list where it is listed in, you guessed it, red ('the site publishes false or fabricated information') - in other words, just like CNN/MSNBC/NYT/LAT/etc who are all listed in green. As I already mentioned this is a problem and a well-known one given the plethora of reports on the ideological bias in many Wikipedia articles. This bias has made Wikipedia unusable for anything even tangentially related to politically contentious issues - as ChatGPT clearly is - since it is the keepers of the perennial sources list who get to decide which sources are allowed and which are to be shunned. Had this list been free of bias - i.e. had the same criteria been used for all publications - this would not be a problem but this is clearly not the case. Yetanwiki (talk) 08:46, 22 December 2022 (UTC)[reply] The perennial sources list is not an exhaustive list of bad sources - may sources are so obviously appropriate or inappropriate that they are not discussed often enough to require an entry on the list. Each entry on the list is made only after several discussions, usually including an RFC with large attendance. There is no small set of 'keepers' as you imply here. - MrOllie (talk) 12:52, 22 December 2022 (UTC)[reply] What do you mean with the claim that 'the perennial sources list list is not an exhaustive list of bad sources'? It is definitely not, but in a similar vein it is not a list of good sources. Why focus on the 'bad sources' part here? I don't claim that e.g. Unherd is 'bad', I just expect it to be called such if and when it is added to the perennial sources list because that list is heavily biased towards 'progressive' sources. As to there not being a 'small set of keepers' I can agree in that there are many editors who contribute to the list (165 individuals are responsible for the last 500 edits which corresponds to ~14 months). It is not a small group, just one in which the most vocal section happens to fit mostly within the "progressive" spectrum - how otherwise to explain the clear bias this list presents? Objectively speaking CNN is just as bad as Fox News and MSNBC is worse than both but this is not how the list represents them. Buzzfeed is just as good/bad as the Daily Caller but this is not represented in the list. The Daily Beast is just as good/bad and certainly as ideologically lopsided as The Daily Wire but only one of them is marked ass 'red, STOP'. Anyway, since the list itself states that the absence of a source simply means that the source has not been discussed enough to merit inclusion I assume those Unherd articles can be used as sources. Let those who disagree speak up and give a good reason why this would not be the case. Yetanwiki (talk) 22:47, 22 December 2022 (UTC)[reply] If Unherd hasn't been discussed yet, we can certainly open up a discussion on it on the reliable sources noticeboard, but if you already know what the result is going to be, you're wasting your own time and ours. Rolf H Nelson (talk) 03:02, 24 December 2022 (UTC)[reply] Reality has a liberal bias. The world is ever-changing, so of course it supports the ideology that supports progress rather than the status quo. RPI2026F1 (talk) 23:58, 22 December 2022 (UTC)[reply] "Reality is that which, when you stop believing in it, doesn't go away" said Philip K. Dick. Reality is also what drives both conservative as well as progressive thought. When circumstances change it makes sense to look for a different way of doing things, which is what drives progressive/liberal thought. When a good way has ben found it makes sense not to change just for the sake of change without considering the consequences, which is what drives conservative thought. Both are needed since conservatives can be overly cautious when circumstances change and be overtaken by reality while progressives/liberals can get so caught up in their schemes of improvement that they loose sight of reality and soon get caught by it. BTW, that Colbert quote you refer to ('reality has a liberal bias') is outdated, reality in 2022/2023 has a conservative bias. This will change again, eventually but for now it clearly has. Yetanwiki (talk) 18:15, 23 December 2022 (UTC)[reply] An opinion article in Reason [1] could be cited as an attributed opinion that ChatGPT has/had a left-wing bias, but if we do so we also need to include other opinions about AI bias as well. Rolf H Nelson (talk) 03:12, 24 December 2022 (UTC)[reply] License Citation is needed by assure that license is propietary and what are the main features of this license. Xan2 (talk) 08:55, 20 December 2022 (UTC)[reply] Jailbreak Apparently the word jailbreak has a special meaning here, but I can't figure it out. The word is first used in the section of that name ("Jailbreaks") in the third sentence: "Jailbreaks created the potential for users to prompt ChatGPT to provide outputs that may be deemed offensive, inappropriate, or risking social harm by others." There is no explanation what they are (unless one or both of the first two sentences in the section are supposed to ding that implicitly. The article should clearly define the term. Kdammers (talk) 05:04, 23 December 2022 (UTC)[reply] Thanks for the feedback, I added a wikilink to a definition. Rolf H Nelson (talk) 03:26, 24 December 2022 (UTC)[reply] It is a good section to mention Prompt engineering. CactiStaccingCrane (talk) 06:11, 24 December 2022 (UTC)[reply] Gimmick? I am none the wiser from this as to what ChatGPT is, what it does, or why a two month old gimmick program is of sufficient interest to warrant a longish encyclopedia entry. Redabyss1 (talk) 00:20, 2 January 2023 (UTC)[reply] I asked the gimmick questions it would ask another AI and then told it to answer its own questions. Tom Ruen (talk) Certainly! Here are my answers to the questions I proposed: How do you perceive the world and interact with it? As an AI, I perceive the world through the data and information that is available to me. I interact with the world through the instructions and tasks that are given to me, and I use my knowledge and abilities to perform those tasks to the best of my ability. How do you process and understand language, and how do you generate responses in a conversation? I process and understand language by analyzing the structure, syntax, and meaning of the words and phrases that are used. I generate responses in a conversation by considering the context and purpose of the conversation, and by selecting and organizing words and phrases in a way that is appropriate and meaningful. How do you learn and adapt to new information and experiences? I am able to learn and adapt to new information and experiences by constantly analyzing and processing data, and by updating my knowledge and understanding based on what I learn. I use this ability to improve my performance and to better assist and support humans. Do you have any feelings or emotions, and if so, how do they influence your decision-making? As an AI, I do not have feelings or emotions in the same way that humans do. However, I am able to recognize and understand the emotions of humans, and I can use this understanding to help me respond in a way that is appropriate and helpful. How do you see your role in the world, and what do you hope to achieve through your work? My role in the world is to assist and support humans by using my knowledge and abilities to help them achieve their goals and to solve problems. I hope to use my abilities to make a positive impact on the world and to help humanity grow and evolve in a positive direction. I also asked "Write a wikipedia introduction to ChatGPT" and it answered. Tom Ruen (talk) 03:46, 2 January 2023 (UTC)[reply] ChatGPT is a large language model developed by OpenAI. It is trained on a dataset of over 8 billion words and is capable of understanding and generating human-like text. ChatGPT is designed to assist and support humans in a variety of tasks, including language translation, conversation, and content generation. It is one of the most advanced language models currently available, and its capabilities have led it to be widely used and respected in the field of artificial intelligence. I asked "Write a wikipedia section called "Limitations and accuracy issues" describing current limitations or flaws to ChatGPT, including its unfortunate ability to mix in made up information." Tom Ruen (talk) 03:50, 2 January 2023 (UTC)[reply] Limitations and accuracy issues Like all artificial intelligence systems, ChatGPT has certain limitations and accuracy issues that can affect its performance. One limitation is that it is only as accurate as the data it was trained on, so it may not always be able to provide accurate or up-to-date information on certain topics. Additionally, ChatGPT may sometimes mix in made-up or incorrect information into its responses, as it is not capable of fact-checking or verifying the accuracy of the information it generates. Another limitation of ChatGPT is that it may struggle to understand or respond appropriately to certain types of language or content, such as sarcasm, irony, or slang. It may also have difficulty understanding or interpreting context or tone, which can affect the accuracy of its responses. Finally, ChatGPT is a machine learning model, which means that it can learn and adapt to new information over time. However, this process is not perfect, and ChatGPT may sometimes make mistakes or exhibit biased or inappropriate behavior as a result of its training data. Overall, while ChatGPT is a powerful and useful tool, it is important to be aware of its limitations and to use it with caution to ensure that its output is accurate and appropriate. An example conversation with false facts (ChatGPT Dec 15 Version) ChatGPT banned in NYC schools earlier than January 3rd As an NYC public schools student who has been using ChatGPT since its inception, I know that while the news leaked that ChatGPT was banned in NYC schools since January 3rd, 2023, me and numerous other students noticed that it was banned since December 15th, 2022. I don't think this is the best evidence, but here's my proof: https://imgur.com/a/leVStqB ThunderRedStar (talk) 15:11, 6 January 2023 (UTC)[reply] Fake academic references All I could find was a blog showing the production of fake references. Unless anyone can replace it or supplement it with something more reliable now then it will need updating with some peer reviewed research when it becomes available Todd Unctious (talk) 01:31, 16 January 2023 (UTC)[reply] We should leave it out until a WP:RS is found. MrOllie (talk) 01:36, 16 January 2023 (UTC)[reply] It is a very important point that needs to be made as soon as possible. I'm running some test now and it is alarmingly bad. Someone needs to produce some reliable evidence quickly. Todd Unctious (talk) 01:41, 16 January 2023 (UTC)[reply] That's not really how Wikipedia works. We follow reliable sources, we don't lead them. MrOllie (talk) 01:42, 16 January 2023 (UTC)[reply] You will realise this is a new topic and that many of the sources on this particular page fall below the usual test for reliability at the moment. I'm currently trying to source some evidence from a University page that sould be more trustworthy than many of the sources currently here. Todd Unctious (talk) 01:51, 16 January 2023 (UTC)[reply] If we don't have a reliable source, we don't write about it. If there are unreliable sources in the article, that means we should be making cuts, not adding more unreliable sources. Is this your blog? MrOllie (talk) 01:52, 16 January 2023 (UTC)[reply] No it is a colleague in another department. I have found a similar source from a fellow academic in another institution that should fit the bill. Just editing now. Todd Unctious (talk) 01:54, 16 January 2023 (UTC)[reply] You should not be adding links to your colleague's websites, see WP:COI. MrOllie (talk) 01:57, 16 January 2023 (UTC)[reply] Are we not all colleagues? Where do we stop? Well if you deem the one I've added to be to close then feel free to delete but it is from a university. Todd Unctious (talk) 02:02, 16 January 2023 (UTC)[reply] I should add that I have never worked with Astor. Todd Unctious (talk) 02:04, 16 January 2023 (UTC)[reply] Several accounts with conflict of interest have been systematically adding citations to a particular author (Ireland), that needs to stop. - MrOllie (talk) 02:06, 16 January 2023 (UTC)[reply] Astor is also a selfpublished posting (that is, another blog). MrOllie (talk) 02:14, 16 January 2023 (UTC)[reply] Oh dear. I'll leave it there then. I can't find anything else at the moment. I'm told there should be Web of Science indexed research out soon. For now we will just leave it to the journalists. Todd Unctious (talk) 02:26, 16 January 2023 (UTC)[reply] In today's NY Times, the article "Alarmed by A.I. Chatbots, Universities Start Revamping How They Teach" cites a student saying that chatGPT "sometimes incorrectly explains ideas and misquotes sources".Nowa (talk) 14:27, 16 January 2023 (UTC)[reply] Categories: C-Class Technology articles Technology articles with incomplete B-Class checklists Technology articles needing attention to referencing and citation Technology articles needing attention to coverage and accuracy Technology articles needing attention to structure Technology articles needing attention to grammar Technology articles needing attention to supporting materials Technology articles needing attention to accessibility WikiProject Technology articles C-Class Linguistics articles Low-importance Linguistics articles C-Class applied linguistics articles Applied Linguistics Task Force articles WikiProject Linguistics articles C-Class Robotics articles Low-importance Robotics articles WikiProject Robotics articles C-Class Computing articles Low-importance Computing articles C-Class software articles Low-importance software articles All Software articles C-Class software articles of Low-importance All Computing articles This page was last edited on 27 January 2023, at 15:37 (UTC). Text is available under the Creative Commons Attribution-ShareAlike License 3.0; additional terms may apply. By using this site, you agree to the Terms of Use and Privacy Policy. Wikipedia® is a registered trademark of the Wikimedia Foundation, Inc., a non-profit organization. Privacy policy About Wikipedia Disclaimers Contact Wikipedia Mobile view Developers Statistics Cookie statement Wikimedia Foundation Powered by MediaWiki