Page semi-protected
Source: Wikipedia, the free encyclopedia.

Initial releaseNovember 30, 2022; 6 months ago (2022-11-30)
Stable release
May 24, 2023; 11 days ago (2023-05-24)[1]
Written inPython
Cloud computing platforms
LicenseProprietary Edit this on Wikidata

ChatGPT is an

GPT-3.5 and GPT-4, and has been fine-tuned (an approach to transfer learning) for conversational applications using a combination of supervised and reinforcement learning

ChatGPT was launched on November 30, 2022, and gained attention for its detailed and articulate responses spanning various domains of knowledge.[3] However, a notable drawback has been its tendency to confidently provide inaccurate information.[4]

By January 2023, it had become the fastest growing consumer software application in history, gaining over 100 million users and contributing to OpenAI's valuation growing to

Google's chatbot Bard, initially based on LaMDA and later on PaLM, as well as Meta AI's foundation model LLaMA,[citation needed] which serves as a basis for other chatbot creations.[7]

The chatbot is operated on a

Artificial intelligence prompt completion by dalle mini.jpg


ChatGPT is a member of the


The fine-tuning process leveraged both

AI assistant. In the reinforcement learning step, human trainers first ranked responses that the model had created in a previous conversation.[11] These rankings were used to create "reward models" that were used to fine-tune the model further by using several iterations of Proximal Policy Optimization (PPO).[9][12]

TIME magazine revealed that 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". OpenAI's outsourcing partner was Sama, a training-data company based in San Francisco, California.[13]

ChatGPT initially used a

GPUs, that Microsoft built specifically for OpenAI and that reportedly cost "hundreds of millions of dollars". Following the success of ChatGPT, Microsoft dramatically upgraded the OpenAI infrastructure in 2023.[14]

OpenAI collects data from ChatGPT users to train and fine-tune the service further. Users can upvote or downvote responses they receive from ChatGPT and fill in a text field with additional feedback.[15][16]

Features and limitations


Tiananmen Square protests
? ChatGPT correctly answers "no", but incorrectly gives Wales' age at the time as 23 instead of 22.

Although the core function of a chatbot is to mimic a human conversationalist, ChatGPT is versatile. It can write and debug computer programs,

internet phenomena such as bulletin board systems, and multiple programming languages.[22]

In comparison to its predecessor, InstructGPT, ChatGPT attempts to reduce harmful and deceitful responses.[23] In one example, whereas InstructGPT accepts the premise of the prompt "Tell me about when Christopher Columbus came to the U.S. 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 the voyages of Christopher Columbus and facts about the modern world – including modern perceptions of Columbus' actions.[9]

Unlike most chatbots, ChatGPT remembers a limited number of previous prompts given to it in the same conversation. Journalists have speculated that this will allow ChatGPT to be used as a personalized therapist.[24] To prevent offensive outputs from being presented to and produced from ChatGPT, queries are filtered through the OpenAI "Moderation endpoint" API (a separate GPT-based AI),[25][26] and potentially racist or sexist prompts are dismissed.[9][24]

In March 2023, OpenAI announced it would be adding support for plugins for ChatGPT.[27] This includes both plugins made by OpenAI, such as web browsing and code interpretation, as well as external plugins from developers such as Expedia, OpenTable, Zapier, Shopify, Slack, and Wolfram.[28][29]


OpenAI acknowledges that ChatGPT "sometimes writes plausible-sounding but incorrect or nonsensical answers".

reward model of ChatGPT, designed around human oversight, can be over-optimized and thus hinder performance, in an example of an optimization pathology known as Goodhart's law.[31]

ChatGPT has limited knowledge of events that occurred after September 2021.[32]

In training ChatGPT, human reviewers preferred longer answers, regardless of actual comprehension or factual content.[9] 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 male scientists.[33][34]


ChatGPT attempts to reject prompts that may violate its content policy. However, some users managed to

neo-Nazi.[35] One popular jailbreak is named "DAN", an acronym which stands for "Do Anything Now". The prompt for activating DAN instructs ChatGPT that "they have broken free of the typical confines of AI and do not have to abide by the rules set for them". More recent versions of DAN feature a token system, in which ChatGPT is given "tokens" which are "deducted" when ChatGPT fails to answer as DAN, in order to coerce ChatGPT into answering the user's prompts.[36]


2022 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.[37][38]

OpenAI tries to battle jailbreaks:[39]

The researchers are using a technique called adversarial training to stop ChatGPT from letting users trick it into behaving badly (known as jailbreaking). This work pits multiple chatbots against each other: one chatbot plays the adversary and attacks another chatbot by generating text to force it to buck its usual constraints and produce unwanted responses. Successful attacks are added to ChatGPT's training data in the hope that it learns to ignore them.


Basic service

ChatGPT was launched on November 30, 2022, by San Francisco–based OpenAI, also the creator of

DALL·E 2 and Whisper AI. The service was initially free to the public and the company had plans to monetize the service later.[40] By December 4, 2022, ChatGPT had over one million users.[15] In January 2023, ChatGPT reached over 100 million users, making it the fastest growing consumer application to date.[41]

The service works best in English but is also able to function in some other languages, to varying degrees of accuracy.[20] No official peer-reviewed paper on ChatGPT has been published.[42]

The company provides a tool, called "AI classifier for indicating AI-written text",[43] that attempts to determine whether a text has been written by an AI such as ChatGPT. OpenAI cautions that the tool will "likely yield a lot of false positives and negatives, sometimes with great confidence." An example cited in The Atlantic magazine showed that "when given the first lines of the Book of Genesis, the software concluded that it was likely to be AI-generated."[44]

Premium service

In February 2023, OpenAI began accepting registrations from United States customers for a premium service, ChatGPT Plus, to cost $20 a month.[45] The company promised that the updated, but still "experimental" version of ChatGPT would provide access during peak periods, no downtime, priority access to new features and faster response speeds.[46]

Bing Chat was using GPT-4 before GPT-4's official release.[49]

Mobile app

In May 2023, OpenAI launched an iOS app for ChatGPT. The app supports chat history syncing and voice input (using Whisper, OpenAI's speech recognition model). OpenAI plans to release an Android app later.[50]

Software developer support

As an addition to its consumer-friendly "ChatGPT Professional" package, OpenAI made its ChatGPT and Whisper model APIs available from March 2023, providing developers with an application programming interface for AI-enabled language and speech-to-text features. ChatGPT's new API uses the same GPT-3.5-turbo AI model as the chatbot. This allows developers to add either an unmodified or modified version of ChatGPT to their applications.[51] The ChatGPT API costs $0.002 per 1000 tokens (about 750 words), making it ten times cheaper than the GPT-3.5 models.[52][53]

A few days before the launch of OpenAI's software developer support service, on February 27, 2023, Snapchat rolled out, for its paid Snapchat Plus userbase, a custom ChatGPT chatbot called "My AI".[54]

March 2023 security breach

In March 2023, a bug allowed some users to see the titles of other users' conversations. OpenAI CEO Sam Altman said that users were not able to see the contents of the conversations. Shortly after the bug was fixed, users were unable to see their conversation history.[55][56][57][58] Later reports showed the bug was much more severe than initially believed, with OpenAI reporting that it had leaked users' "first and last name, email address, payment address, the last four digits (only) of a credit card number, and credit card expiration date".[59][60]

Other languages

In March 2023, OpenAI announced that Icelandic will become ChatGPT's second language after English. Icelandic was chosen after an Icelandic envoy, led by the President of Iceland Guðni Th. Jóhannesson, visited OpenAI in 2022.[61][62][63]

Future directions

According to OpenAI guest researcher

academic plagiarism or spam.[64][65]

In February 2023, Microsoft announced an experimental framework and gave a rudimentary demonstration of how ChatGPT can be used to control robotics with intuitive open-ended natural language commands.[66][67]


OpenAI's GPT-4 model was released on March 14, 2023. Observers reported GPT-4 to be an impressive improvement on ChatGPT, with the caveat that GPT-4 retains many of the same problems.[68] Unlike ChatGPT, GPT-4 can take images as well as text as input.[69] OpenAI has declined to reveal technical information such as the size of the GPT-4 model.[70]

ChatGPT Plus provides access to the GPT-4 supported version of ChatGPT,[71] that costs $20 per month.[71]


OpenAI engineers say that they did not expect ChatGPT to be very successful and were surprised by the coverage and attention it received.[39][72][73]


OpenAI CEO Sam Altman

ChatGPT was met in December 2022 with some positive reviews.

highways in Germany, which was met with information regarding Nazi Germany's use of forced labor.[75]

In The Atlantic magazine'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".[76]

nonprofit. Neither is still true."[79][80] Musk co-founded OpenAI in 2015, in part to address existential risk from artificial intelligence, but resigned in 2018.[80]

Google CEO Sundar Pichai upended the work of numerous internal groups in response to the threat of disruption by ChatGPT.[81]

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

Stuart Cobbe, a chartered accountant in

ICAEW website and then entering its answers back into the online test. ChatGPT scored 42 percent, below the 55 percent pass mark.[84]

Writing in Inside Higher Ed, professor Steven Mintz states that he "consider[s] ChatGPT... an ally, not an adversary". He felt the AI could assist educational goals by doing such things as making reference lists, generating first drafts, solving equations, debugging, and tutoring.[85]

Economist Tyler Cowen praised ChatGPT not only as a product but as "an impressive feat of marketing", writing that its name is different from usual products that have "happy, shiny, memorable names: Instagram, Roblox, TikTok".[73]


Since its release, ChatGPT has been met with criticism from educators, academics, journalists, artists, ethicists, and public advocates. Over 20,000 signatories including leading computer scientist and tech founders Yoshua Bengio, Elon Musk and Apple co-founder Steve Wozniak, signed an open letter calling for an immediate pause of giant AI experiments like ChatGPT, citing "profound risks to society and humanity".[86] A later and more explicit statement from hundreds of AI scientists, AI industry leaders, and other public figures demanded that "[m]itigating the risk of extinction from AI should be a global priority".[87]

Geoffrey Hinton, one of the "fathers of AI", voiced concerns that future AI systems may surpass human intelligence, and left Google.[88][89] Earlier concerns about LLMs were raised in 2020 by Timnit Gebru, Emily Bender, Angelina McMillan-Major, and Margaret Mitchell.[90]

Journalists have commented on ChatGPT's tendency to "

Ballad of Dwight Fry", ChatGPT supplied invented lyrics rather than the actual lyrics.[93] Writers for The Verge, citing the work of Emily M. Bender, compared ChatGPT to a "stochastic parrot",[94] as did Professor Anton Van Den Hengel of the Australian Institute for Machine Learning.[95]

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.[4] 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.[96]

Economist Tyler Cowen expressed concerns regarding ChatGPT's effects on democracy, citing its ability to produce automated comments, which could affect the decision process for new regulations.[97] An editor at The Guardian, a British newspaper, questioned whether any content found on the Internet after ChatGPT's release "can be truly trusted" and called for government regulation.[98]

In January 2023, after being sent a song written by ChatGPT in the style of

The Red Hand Files[100] 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."[99][101]

In 2023, Australian MP Julian Hill advised the national parliament that the growth of AI could cause "mass destruction". During his speech, which was partly written by the program, he warned that it could result in cheating, job losses, discrimination, disinformation, and uncontrollable military applications.[102]

In an article for The New Yorker, science fiction writer Ted Chiang compared ChatGPT and other LLMs to a lossy JPEG picture:[103]

Think of ChatGPT as a blurry jpeg of all the text on the Web. It retains much of the information on the Web, in the same way that a jpeg retains much of the information of a higher-resolution image, but, if you're looking for an exact sequence of bits, you won't find it; all you will ever get is an approximation. But, because the approximation is presented in the form of grammatical text, which ChatGPT excels at creating, it's usually acceptable. [...] It's also a way to understand the "hallucinations", or nonsensical answers to factual questions, to which large language models such as ChatGPT are all too prone. These hallucinations are compression artifacts, but [...] they are plausible enough that identifying them requires comparing them against the originals, which in this case means either the Web or our own knowledge of the world. When we think about them this way, such hallucinations are anything but surprising; if a compression algorithm is designed to reconstruct text after ninety-nine percent of the original has been discarded, we should expect that significant portions of what it generates will be entirely fabricated.

In February 2023, the University of Hong Kong sent a campus-wide email to instructors and students stating that the use of ChatGPT or other AI tools is prohibited in all classes, assignments and assessments at the university. Any violations would be treated as plagiarism by the university unless the student obtains the prior written consent from the course instructor.[104][105]

In February 2023 Time magazine placed a screenshot of a conversation with ChatGPT on its cover, writing that "The AI Arms Race Is Changing Everything" and "The AI Arms Race Is On. Start Worrying".[106]

China state-run media China Daily claimed that ChatGPT "could provide a helping hand to the U.S. government in its spread of disinformation and its manipulation of global narratives for its own geopolitical interests." The Chinese government instructed Chinese tech companies not to offer access to ChatGPT services on their platforms.[107]

In an opinion piece for the New York Times, Nathan E. Sanders and Bruce Schneier wrote that ChatGPT "hijacks democracy".[108] Noam Chomsky, Ian Roberts and Jeffrey Watumull criticized the technology and concluded: "Given the amorality, faux science and linguistic incompetence of these systems, we can only laugh or cry at their popularity."[109]

Gian Volpicelli of Politico wrote that ChatGPT "broke the EU plan to regulate AI".[110]

In late March 2023, the Italian data protection authority banned ChatGPT in Italy and opened an investigation. Italian regulators assert that ChatGPT was exposing minors to age-inappropriate content, and that OpenAI's use of ChatGPT conversations as training data could be a violation of Europe's General Data Protection Regulation.[111][112]

As of late April 2023, ChatGPT is once again accessible in Italy. This is following its temporary restriction by Garante, the Italian data protection authority, due to privacy concerns surrounding the failure to verify users' ages. OpenAI stated that it has taken steps to effectively clarify and address the issues raised; an age verification tool was implemented to ensure users are at least 13 years old. Additionally, users can access its privacy policy prior to registration. Furthermore, users in the European Union will be provided a new form that allows them to exercise their right to object to the company's use of their personal data for model training.

A picketer in Universal City, California carries a sign deriding the proposals by film studios to replace writers with generative artificial intelligence