You can edit almost every page by Creating an account. Otherwise, see the FAQ.

ChatGPT and Wikipedia

From EverybodyWiki Bios & Wiki




ChatGPT
Wikipedia

ChatGPT and Wikipedia are two huge, universal, non-independent sources of knowledge. Their mutual interdependence has led to their cooperation in a short time and in several areas.

Wikipedia is essential for the training, operation and development of ChatGPT,[1] and as their partnership has gradually developed, their relationship has become increasingly close and interactive. However, the possibilities and limitations of using ChatGPT Wikipedia at editorial level have yet to be defined at an international level.[2]

Wikipedia as a contributor to ChatGPT[edit]

ChatGPT (Generative Pre-trained Transformer)[3] is an artificial intelligence model that generates human-like text. The Harvard Business Review hailed the launch of ChatGPT as the dawn of a new era of artificial intelligence (AI).[4] When a user types a question, command or comment into the ChatGPT engine's dialog box, the system almost instantly provides a text response in the same language.[5][6]

The secret of this is that ChatGPT is an artificial intelligence chatbot that uses advanced natural language processing (NLP) as its engine to have realistic conversations with people.[7] At its launch, it was differentiated from other chatbots by its ultra-realistic conversational skills, including the ability to ask follow-up questions, acknowledge mistakes and highlight nuances in topics. Grammatical and syntactical errors were rare, and written structures were logical and understandable.[4]

The chatbot was trained using two types of so-called reinforcement learning methods, RLHF (Reinforcement Learning from Human Feedback)[8][9] and PPO (Proximal Policy Optimization).[10][11] The text input and output were provided by GPT-3. 5 AI algorithm was used (GPT stands for Generative Pre-trained Transformer, which is included in the name of the chatbot itself).[12] GPT-3.5 is an enhanced version of GPT-3, which is a development of OpenAI and uses supervised learning in addition to reinforcement learning.[4]

The training of the model was based on a large amount of human-generated text, mostly web-based text content, which was used for training.[13] NLP does not know the specifics of which documents were used in its training and the weight given to each, nor does it know the specifics of each data source.[14]

It is in this chatbot training phase that the prominent role of Wikipedia in the training became clear,[15] as Wikipedia is an open-source online encyclopedia, a comprehensive collection of human knowledge in various fields, freely accessible to anyone in many languages.[16] Also, for the ChatGPT, which is constantly increasing its knowledge. The content offered by Wikipedia has played a key role in creating the "intelligence" of the AI model, both in quantity and quality.[17] The training and learning process of ChatGPT did not end with its debut in November 2022, but has been ongoing ever since, and this is natural, since, for example, at its premiere it had only a limited knowledge of events after 2021.[18]

As data has become an increasingly dominant factor of power (dataism, Big data),[19][20] Wikipedia has increasingly become a privileged source of facts transformed into data prioritized by the big platforms. Apple's Siri, Microsoft's Bing, Google Search and Home, and Amazon's Alexa have all been seen as such. As early as 2012, Google came up with an innovation that further enhanced the value of Wikipedia, its information and its organized data set.[21] Google's new brain center was called the Knowledge graph, which housed a huge database of billions of records, mostly extracted from Wikipedia.[21]

One of Wikipedia's most important principles is that it is not the place of first publication, and that articles can only be based on secondary sources, a fact that Wikipedia itself has stated and confirmed through its operational practices, that it does not accept Wikipedia sources from its editors.[22] The reason why Wikipedia was given a special, distinguished role in the training of ChatGPT was that compared to the raw text sources used for other training, it provided more scientifically rigorous, accurate and complex information of a higher quality.[1] This is self-evident, because Wikipedia's articles are subject to continuous and multi-level internal verification, are accessible in over 300 languages, and have a diverse and unique range of content. Of particular importance for ChatGPT's "up-to-dateness" is the fact that its Wikipedia resource is constantly and dynamically updated over time; it provides an ideal platform for further training on a large amount and variety of data.[23] The information provided in Wikipedia's infoboxes is of strategic importance and is treated as a priority by ChatGPT, just as, for example, Google, Bing or Yahoo process information on people, companies, places, events in their knowledge tables or fact boxes.[24]

ChatGPT uses Wikipedia as one of its main resources, but there are limitations to consider.[25] The factuality of Wikipedia is high, but not perfect. ChatGPT must therefore learn to deal with inconsistencies, misinformation and biases.[26] Nevertheless, Wikipedia is "one and only" and any AI tool can inevitably only "train itself" and then function along according to the training principles of ChatGPT.[27] That progressively reinforces and enhances the importance and relevance of Wikipedia's open source, community-driven model.[28]

ChatGPT as a contributor to Wikipedia[edit]

As interdependence grows stronger over time,[29] ChatGPT also offers huge potential for the development of Wikipedia.[30] Artificial intelligence can help with tasks such as detecting vandalism, making editorial suggestions, or even adding to stub articles. These examples show that the relationship can indeed become more symbiotic over time.[24]

ChatGPT can also help to create new content based on existing Wikipedia articles, being able to create simplified versions of complex articles, effectively increasing the accessibility and availability of information.[31] It can improve the reliability and accuracy of Wikipedia, can be used to check the factuality of information, identify and correct biases, and can facilitate the implementation of a neutral point of view (NPOV) for word articles.[32] By improving the quality of content, ChatGPT can contribute to increasing the credibility and thus the relevance of Wikipedia.[27]

ChatGPT can increase the accessibility of Wikipedia by providing translations, summaries for quick understanding, or audio versions of articles. The easier it is for users to access and understand Wikipedia content, the more important and popular Wikipedia becomes as a knowledge resource in general. The above contributes to the promotion of Wikipedia.[33]

Wikipedia community and ChatGPT[edit]

Since the launch of ChatGPT, there has been a heated debate in Wikipedia community about whether and how artificial intelligence can be used in the actual editing of Wikipedia.[34] No recommendations have yet been developed that specifically apply to editing with ChatGPT, although the internal guidelines forbidding plagiarism and requiring verification do, by definition, prohibit blind insertion of AI-generated text.[35] Some Wikipedians argue that ChatGPT should be banned altogether, even if the articles so produced were later checked by live editors, because the AI is too capable of making plausible falsehoods. There would also be a risk that Wikipedia editors would find it harder to police the content posted.[24]

However, as early as 2002, a Wikpedian who edited under the name Ram-Man used an automated tool called rambot to automatically generate Wikipedia pages for US cities based on US census data, thus establishing a whole new editing practice.[2] Partly as a result of this novelty the Wikipedia community in the very same year agreed to allow the use of chatbots in certain cases.[36] Consequently bots that automatically detect vandalism for example and filter out vulgar texts and swearing, are by now widely accepted.[36]

Andrew Lih, a fellow Wikimedian at the Smithsonian Institution in Washington, D.C., who has been a volunteer Wikipedia editor since 2003, argues that the potential of ChatGPT can help a Wikipedian overcome initial inertia and find "activation energy."[2] He sees ChatGPT as a novel way to implement Cunningham's Law.[37] According to Cunningham's Law, the best way to get the right answer on the Internet is not to ask a question, but to post the wrong answer. By analogy, the easiest way now to produce a good page is to publish a rather bad, but not an entirely bad ChatGPT-generated version for further editing.[38]

The first Wikipedia page using ChatGPT was published on December 6, 2022, by Richard Knipel, a long-time Wikipedian who edits under the name Pharos, under the title Artwork title.[2] In his editorial summary, he said that it was just a draft created with ChatGPT, which he would later modify.[39] ChatGPT produced a standard, grammatically correct version on the subject. Wikipedians like Knipel believe that ChatGPT could be used as a tool within Wikipedia without making the human role redundant, as the raw text generated by the chatbot could serve as a useful starting point or outline that could then be checked and elaborated by the editor. Knipel stated: “We always have imperfect information, and then we correct it. If the issue is the original sin of using A.I., well, I don’t believe in original sin”.[2]

References[edit]

  1. 1.0 1.1 Dwivedi, Yogesh K.; Kshetri, Nir; Hughes, Laurie; Slade, Emma Louise; Jeyaraj, Anand; Kar, Arpan Kumar; Baabdullah, Abdullah M.; Koohang, Alex; Raghavan, Vishnupriya; Ahuja, Manju; Albanna, Hanaa; Albashrawi, Mousa Ahmad; Al-Busaidi, Adil S.; Balakrishnan, Janarthanan; Barlette, Yves (2023-08-01). "Opinion Paper: "So what if ChatGPT wrote it?" Multidisciplinary perspectives on opportunities, challenges and implications of generative conversational AI for research, practice and policy". International Journal of Information Management. 71: 102642. doi:10.1016/j.ijinfomgt.2023.102642. ISSN 0268-4012. Archived from the original on 2023-06-24. Retrieved 2023-06-25. Unknown parameter |url-status= ignored (help); Unknown parameter |s2cid= ignored (help)
  2. 2.0 2.1 2.2 2.3 2.4 Harrison, Stephen (2023-01-12). "Should ChatGPT Be Used to Write Wikipedia Articles?". Slate. ISSN 1091-2339. Archived from the original on 2023-01-28. Retrieved 2023-06-25. Unknown parameter |url-status= ignored (help)
  3. Roose, Kevin (2022-12-05). "The Brilliance and Weirdness of ChatGPT". The New York Times. ISSN 0362-4331. Archived from the original on 2023-01-18. Retrieved 2023-06-25. Unknown parameter |url-status= ignored (help)
  4. 4.0 4.1 4.2 Greengard, Samuel (2022-12-29). "ChatGPT: Understanding the ChatGPT AI Chatbot". eWEEK. Archived from the original on 2023-01-19. Retrieved 2023-06-25. Unknown parameter |url-status= ignored (help)
  5. Lock, Samantha (2022-12-05). "What is AI chatbot phenomenon ChatGPT and could it replace humans?". The Guardian. ISSN 0261-3077. Archived from the original on 2023-01-16. Retrieved 2023-06-27. Unknown parameter |url-status= ignored (help)
  6. "What is ChatGPT, the artificial intelligence text bot that went viral?". ABC News. Archived from the original on 2023-06-27. Retrieved 2023-06-27. Unknown parameter |url-status= ignored (help)
  7. McFadden, Christopher (2023-02-01). "ChatGPT: 30 incredible ways to use the AI-powered chatbot". interestingengineering.com. Retrieved 2023-06-27.
  8. "Learning from human preferences". openai.com. Archived from the original on 2023-06-18. Retrieved 2023-06-25. Unknown parameter |url-status= ignored (help)
  9. Deep reinforcement learning from human preferences Paul Christiano, Jan Leike, Tom B. Brown, Miljan Martic, Shane Legg, Dario Amodei --- [Submitted on 12 Jun 2017 (v1), last revised 17 Feb 2023 (this version, v4)] https://arxiv.org/abs/1706.03741v4 Archived 2023-04-29 at the Wayback Machine
  10. "Proximal Policy Optimization". openai.com. Archived from the original on 2023-06-24. Retrieved 2023-06-25. Unknown parameter |url-status= ignored (help)
  11. Proximal Policy Optimization Algorithms John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, Oleg Klimov --- [Submitted on 20 Jul 2017 (v1), last revised 28 Aug 2017 (this version, v2)] https://arxiv.org/abs/1707.06347v2 Archived 2023-06-24 at the Wayback Machine
  12. "What is ChatGPT, DALL-E, and generative AI? | McKinsey". www.mckinsey.com. Archived from the original on 2023-04-23. Retrieved 2023-06-27. Unknown parameter |url-status= ignored (help)
  13. "How ChatGPT and similar AI will disrupt education". 2023-04-12. Archived from the original on 2023-06-01. Retrieved 2023-06-27. Unknown parameter |url-status= ignored (help)
  14. "Machine learning, explained | MIT Sloan". mitsloan.mit.edu. 2023-06-20. Archived from the original on 2023-06-23. Retrieved 2023-06-25. Unknown parameter |url-status= ignored (help)
  15. "The inside story of ChatGPT: How OpenAI founder Sam Altman built the world's hottest technology with billions from Microsoft". Fortune. Archived from the original on 2023-05-02. Retrieved 2023-06-27. Unknown parameter |url-status= ignored (help)
  16. "An AI Helped Write This Play. It May Contain Racism". Time. 2021-08-23. Archived from the original on 2023-05-23. Retrieved 2023-06-27. Unknown parameter |url-status= ignored (help)
  17. Montti, Roger (2022-12-12). "ChatGPT For Content and SEO?". Search Engine Journal. Archived from the original on 2023-03-09. Retrieved 2023-06-25. Unknown parameter |url-status= ignored (help)
  18. "Could a chatbot answer Prime Minister's Questions?". BBC News. 2022-12-27. Archived from the original on 2023-01-17. Retrieved 2023-06-25. Unknown parameter |url-status= ignored (help)
  19. Adams, Tim (2016-09-11). "Homo Deus: A Brief History of Tomorrow by Yuval Noah Harari review – chilling". The Observer. ISSN 0029-7712. Archived from the original on 2023-06-03. Retrieved 2023-06-27. Unknown parameter |url-status= ignored (help)
  20. Harari, Yuval Noah (2016-08-26). "Yuval Noah Harari on big data, Google and the end of free will". Financial Times. Archived from the original on 2019-08-22. Retrieved 2023-06-27. Unknown parameter |url-status= ignored (help)
  21. 21.0 21.1 "What is a Knowledge Graph? A comprehensive Guide". WordLift Blog. Archived from the original on 2023-06-25. Retrieved 2023-06-25. Unknown parameter |url-status= ignored (help)
  22. "Wikipédia szövegírás, SEO Wikipédia készítés és tanácsadás". Tartalomkeszites.hu (in magyar). Archived from the original on 2023-03-27. Retrieved 2023-06-25. Unknown parameter |url-status= ignored (help)
  23. Perez, Sarah (2023-01-18). "Wikipedia gets its first makeover in over a decade... and it's fairly subtle". TechCrunch. Archived from the original on 2023-01-19. Retrieved 2023-06-28. Unknown parameter |url-status= ignored (help)
  24. 24.0 24.1 24.2 Ford, Heather. "Writing the Revolution: Wikipedia and the Survival of Facts in the Digital Age". Next Big Idea Club. Archived from the original on 2023-06-24. Retrieved 2023-06-25. Unknown parameter |url-status= ignored (help)
  25. "Wikipedia's 20, but how credible is it? – DW – 01/14/2021". dw.com. Archived from the original on 2021-01-19. Retrieved 2023-06-28. Unknown parameter |url-status= ignored (help)
  26. Jeremy Hsu (2009-11-06). "Wikipedia: How Accurate Is It?". livescience.com. Archived from the original on 2023-03-26. Retrieved 2023-06-28. Unknown parameter |url-status= ignored (help)
  27. 27.0 27.1 Saad, Nichole (2023-01-19). "For educators, is ChatGPT the new Wikipedia?". Medium. Archived from the original on 2023-06-24. Retrieved 2023-06-25. Unknown parameter |url-status= ignored (help)
  28. Naughton, John (2014-08-10). "Wikipedia isn't perfect, but as a model it's as good as it gets". The Guardian. ISSN 0261-3077. Archived from the original on 2022-07-02. Retrieved 2023-06-28. Unknown parameter |url-status= ignored (help)
  29. Harari, Yuval Noah (2016-08-26). "Yuval Noah Harari on big data, Google and the end of free will". Financial Times. Archived from the original on 2019-08-22. Retrieved 2023-06-28. Unknown parameter |url-status= ignored (help)
  30. Jones, Eric. "Council Post: The Strategic Opportunities Of Advanced AI: A Focus On ChatGPT". Forbes. Archived from the original on 2023-05-22. Retrieved 2023-06-28. Unknown parameter |url-status= ignored (help)
  31. "9 Ways That ChatGPT Can Help Content Creators". MUO. 2023-06-20. Archived from the original on 2023-06-25. Retrieved 2023-06-28. Unknown parameter |url-status= ignored (help)
  32. "What is ChatGPT? Everything you need to know about chatbot from OpenAI". Washington Post. 2022-12-06. Archived from the original on 2023-03-29. Retrieved 2023-06-28. Unknown parameter |url-status= ignored (help)
  33. "ChatGPT use cases: 12 examples to use AI effectively". Dexerto. Archived from the original on 2023-06-24. Retrieved 2023-06-25. Unknown parameter |url-status= ignored (help)
  34. Saad, Nichole (2023-01-19). "For educators, is ChatGPT the new Wikipedia?". Medium. Archived from the original on 2023-06-24. Retrieved 2023-06-28. Unknown parameter |url-status= ignored (help)
  35. "Wikipedia Uses iThenticate to Ensure Content Quality". www.turnitin.com. 2019-05-14. Archived from the original on 2022-11-29. Retrieved 2023-06-28. Unknown parameter |url-status= ignored (help)
  36. 36.0 36.1 Vandalism Detection in Wikipedia: a Bag-of-Words Classifier Approach --- Amit Belani --- ab422@cornell.edu November 11, 2009 https://arxiv.org/ftp/arxiv/papers/1001/1001.0700.pdf Archived 2023-04-02 at the Wayback Machine
  37. "What 'Cunningham's Law' Really Tells Us About How We Interact Online". Lifehacker. 2022-04-01. Archived from the original on 2023-06-07. Retrieved 2023-06-28. Unknown parameter |url-status= ignored (help)
  38. "Cunningham's Law – Brian.Carnell.Com". brian.carnell.com. Archived from the original on 2022-12-01. Retrieved 2023-06-28. Unknown parameter |url-status= ignored (help)
  39. "Artwork title: Revision history - Wikipedia". en.wikipedia.org. Archived from the original on 2023-01-17. Retrieved 2023-06-28. Unknown parameter |url-status= ignored (help)

{{#set:Technical tag=Article from Wikipedia}}{{#set:priority= }} {{#set:PageName=ChatGPT_and_Wikipedia }}

This article "ChatGPT and Wikipedia" is from Wikipedia. The list of its authors can be seen in its historical and/or the page Edithistory:ChatGPT and Wikipedia. Articles copied from Draft Namespace on Wikipedia could be seen on the Draft Namespace of Wikipedia and not main one.{{#set:Article=true}}