ChatGPT: Difference between revisions

From AI Wiki
No edit summary
m (Text replacement - "|List of ChatGPT Plugins" to "|Custom GPTs")
 
(26 intermediate revisions by 4 users not shown)
Line 1: Line 1:
{{see also|ChatGPT Guides}}
{{see also|Custom GPTs|ChatGPT Guides}}
==Introduction==
==Introduction==
[[File:Gpt4-122.jpg|thumb|Figure 1: ChatGPT user interface. Source: OpenAI.]]
[[File:Gpt4-122.jpg|thumb|Figure 1: ChatGPT user interface. Source: OpenAI.]]
Developed by the [[AI]] research lab [[OpenAI]], ChatGPT is a [[large language model]] ([[LLM]]) that generates text based on an [[input]] ([[prompt]]) from a user (figure 1). It interacts in a conversational way, answering follow-up questions, admitting to its mistakes, refusing inappropriate requests, and challenging incorrect premises. <ref name="”1”">OpenAI (2022). ChatGPT: Optimizing language models for dialogue. OpenAI. https://openai.com/blog/chatgpt/</ref> <ref name="”2”">Kelly, SM (2022). This AI chatbot is dominating social media with its frighteningly good essays. CNN International. https://edition.cnn.com/2022/12/05/tech/chatgpt-trnd/index.html</ref> <ref name="”3”">Edwards, B (2022). No Linux? No problem. Just get AI to hallucinate it for you. Ars Technica. https://arstechnica.com/information-technology/2022/12/openais-new-chatbot-can-hallucinate-a-linux-shell-or-calling-a-bbs/</ref> The LLM was trained on a massive amount of information <ref name="”2”" /> and it's a derivative of [[InstructGPT]], a program trained to follow natural language instructions in a prompt and give an elaborate response. <ref name="”1”" />
Developed by the [[AI]] research lab [[OpenAI]], [[ChatGPT]] is a [[large language model]] ([[LLM]]) in the [[GPT|GPT series]] that generates text based on an [[input]] ([[prompt]]) from a user (figure 1). It interacts in a conversational way, answering follow-up questions, admitting to its mistakes, refusing inappropriate requests, and challenging incorrect premises. <ref name="”1”">OpenAI (2022). ChatGPT: Optimizing language models for dialogue. OpenAI. https://openai.com/blog/chatgpt/</ref> <ref name="”2”">Kelly, SM (2022). This AI chatbot is dominating social media with its frighteningly good essays. CNN International. https://edition.cnn.com/2022/12/05/tech/chatgpt-trnd/index.html</ref> <ref name="”3”">Edwards, B (2022). No Linux? No problem. Just get AI to hallucinate it for you. Ars Technica. https://arstechnica.com/information-technology/2022/12/openais-new-chatbot-can-hallucinate-a-linux-shell-or-calling-a-bbs/</ref> The LLM was trained on a massive amount of information <ref name="”2”" /> and it's a derivative of [[InstructGPT]], a program trained to follow natural language instructions in a prompt and give an elaborate response. <ref name="”1”" />


OpenAI's ChatGPT was introduced at the end of November 2022 during a public demo. <ref name="”1”" /> <ref name="”6”">Wiggers, K (2022). While anticipation builds for GPT-4, OpenAI quietly releases GPT-3.5. TechCrunch. https://techcrunch.com/2022/12/01/while-anticipation-builds-for-gpt-4-openai-quietly-releases-gpt-3-5/</ref> In the following days, it was already considered the best [[artificial intelligence]] ([[AI]]) chatbot ever released to the general public. <ref name="”4”">Koose, K (2022). The Brilliance and Weirdness of ChatGPT. The New York Times. https://www.nytimes.com/2022/12/05/technology/chatgpt-ai-twitter.html</ref> Its initial introduction was, according to the AI research lab, "to get users’ feedback and learn about its strengths and weaknesses." They are encouraged to report on problematic model outputs through the UI and false positives and negatives. During this phase, ChatGPT could be used for free at chat.openai.com. <ref name="”1”" />
OpenAI's ChatGPT was introduced at the end of November 2022 during a public demo. <ref name="”1”" /> <ref name="”6”">Wiggers, K (2022). While anticipation builds for GPT-4, OpenAI quietly releases GPT-3.5. TechCrunch. https://techcrunch.com/2022/12/01/while-anticipation-builds-for-gpt-4-openai-quietly-releases-gpt-3-5/</ref> In the following days, it was already considered the best [[artificial intelligence]] ([[AI]]) chatbot ever released to the general public. <ref name="”4”">Koose, K (2022). The Brilliance and Weirdness of ChatGPT. The New York Times. https://www.nytimes.com/2022/12/05/technology/chatgpt-ai-twitter.html</ref> Its initial introduction was, according to the AI research lab, "to get users’ feedback and learn about its strengths and weaknesses." They are encouraged to report on problematic model outputs through the UI and false positives and negatives. During this phase, ChatGPT could be used for free at chat.openai.com. <ref name="”1”" />
Line 20: Line 20:
[[File:GPT Training process.png|thumb|Figure 2: General overview of the training process using reinforcement learning from human feedback. Source: OpenAI.]]
[[File:GPT Training process.png|thumb|Figure 2: General overview of the training process using reinforcement learning from human feedback. Source: OpenAI.]]


The ChatGPT model was trained using [[Reinforcement Learning]] from Human Feedback (RLHF) following the same methods as InstructGPT (figure 2); only with small differences in the data collection setup. Training for an initial model was made using supervised fine-tuning where human AI trainers had "conversations in which they played both sides—the user and an AI assistant." <ref name="”1”" />
The ChatGPT model was trained using [[Reinforcement Learning from Human Feedback]] ([[RLHF]]) following the same methods as InstructGPT (figure 2); only with small differences in the data collection setup. Training for an initial model was made using supervised fine-tuning where human AI trainers had "conversations in which they played both sides—the user and an AI assistant." <ref name="”1”" />


Reinforcement learning uses a reward model for AI training. This was done by collecting comparison data consisting of two or more model responses ranked by quality. According to the official blog of OpenAI, "to collect this data, we took conversations that AI trainers had with the chatbot. We randomly selected a model-written message, sampled several alternative completions, and had AI trainers rank them. Using these reward models, we can fine-tune the model using Proximal Policy Optimization. We performed several iterations of this process." <ref name="”1”" />
Reinforcement learning uses a reward model for [[training|AI training]]. This was done by collecting comparison data consisting of two or more model responses ranked by quality. According to the official blog of OpenAI, "to collect this data, we took conversations that AI trainers had with the chatbot. We randomly selected a model-written message, sampled several alternative completions, and had AI trainers rank them. Using these reward models, we can fine-tune the model using Proximal Policy Optimization. We performed several iterations of this process." <ref name="”1”" />


GPT-3.5 was trained on a mix of text and code published before Q4, 2021. <ref name="”6”" /> Both ChatGPT (a fine-tuned version of a model in the GPT-3.5 series) and GPT-3.5 were trained on an Azure AI supercomputing infrastructure. <ref name="”1”" /> The model on which the chatbot is based, text-davinci-003, can handle more complex instructions with increased output quality and overall better in long-form writing (around 65% longer outputs than text-davinci-002). It also has fewer limitations (e.g. a reduction in "hallucinations") than previous versions and scores higher on human preference rating. <ref name="”6”" />
GPT-3.5 was trained on a mix of text and code published before Q4, 2021. <ref name="”6”" /> Both ChatGPT (a fine-tuned version of a model in the GPT-3.5 series) and GPT-3.5 were trained on an Azure AI supercomputing infrastructure. <ref name="”1”" /> The model on which the chatbot is based, text-davinci-003, can handle more complex instructions with increased output quality and overall better in long-form writing (around 65% longer outputs than text-davinci-002). It also has fewer limitations (e.g. a reduction in "hallucinations") than previous versions and scores higher on human preference rating. <ref name="”6”" />
Line 36: Line 36:
==Applications==
==Applications==


OpenAI's ChatGPT has myriad applications, from generating computer code, to writing jokes, explaining scientific concepts in different levels of complexity, to writing college-level essays. Besides its regular text composition skills, it can also provide assistance, by helping programmers spot and fix errors in code or answer questions made to it, analogous to a Google search. <ref name="”4”" /> Indeed, it was denominated as a possible "Google Killer". <ref name="”9”">Ansari, T (2022). These 8 potential use cases of ChatGPT will blow your mind!​​​​​​​. Analytics India Magazine. https://analyticsindiamag.com/these-8-potential-use-cases-of-chatgpt-will-blow-your-mind/</ref> Due to ChatGPT remembering what a user has written before, some have suggested that it could be possible to create personalized therapy bots. <ref name="”4”" />
OpenAI's ChatGPT has myriad applications, from generating computer code, to writing jokes, explaining scientific concepts in different levels of complexity, to writing college-level essays. Besides its regular text composition skills, it can also provide assistance, by helping programmers spot and fix errors in code or answer questions made to it, analogous to a [[Google]] search. <ref name="”4”" /> Indeed, it was denominated as a possible "Google Killer". <ref name="”9”">Ansari, T (2022). These 8 potential use cases of ChatGPT will blow your mind!​​​​​​​. Analytics India Magazine. https://analyticsindiamag.com/these-8-potential-use-cases-of-chatgpt-will-blow-your-mind/</ref> Due to ChatGPT remembering what a user has written before, some have suggested that it could be possible to create personalized therapy bots. <ref name="”4”" />


Overall, the potential of this application is enormous, possibly having an impact on different business sectors.
Overall, the potential of this application is enormous, possibly having an impact on different business sectors.
*AI prompt generation, for art creation in models like [[DALL-E]] 2 or [[Midjourney]].
*[[AI prompt]] generation, for art creation in models like [[DALL-E]] 2 or [[Midjourney]].
*Act like a virtual cloud, stringing together cloud services to achieve complex tasks.
*Act like a virtual cloud, stringing together cloud services to achieve complex tasks.
*Content generation, changing how content is created for thousands of companies.
*[[Content generation]], changing how content is created for thousands of companies.
*Possible impact on education tech by answering to doubts.
*Possible impact on education tech by answering to doubts.
*Sort, manage, and organize unstructured data.
*Sort, manage, and organize unstructured data.
Line 74: Line 74:
*Faster response times
*Faster response times
*Access to new features and improvements earlier
*Access to new features and improvements earlier
==ChatGPT API==
ChaptGPT API was released on March 1st, 2023 alongside the release of [[Whisper]] API.<ref name="”11”">Introducing ChatGPT and Whisper APIs https://openai.com/blog/introducing-chatgpt-and-whisper-apis</ref>
==ChatGPT Plugins==
{{see also|Custom GPTs}}
[[ChatGPT Plugins]] can extend its capabilities by providing access to recent or specific information not from [[training data]] and allowing it to access external [[APIs]]. There are some plugins created by [[OpenAI]] like [[browsing]], [[code interpretation]] and [[retrieval]]. Developers can also create their own plugins for ChatGPT, and efforts are underway to develop open standards for AI-facing interfaces.
Browsing capabilities, code interpreters, retrieval plugins, and third-party plugins are being explored to expand the range of applications and provide more relevant and up-to-date information for ChatGPT.
==Guides==
{{see also|ChatGPT Guides}}
{{:ChatGPT Guides}}
==Uses==
{{see also|ChatGPT Uses}}
{{:ChatGPT Uses}}
==Updates==
===August 14, 2023===
[[Azure ChatGPT]]
===January 30, 2023===
====Official release note====
We’ve upgraded the ChatGPT model with improved factuality and mathematical capabilities.
====Comments and discoveries====
While ChatGPT became more factual and better at math, before each user entered [[prompts]], [[OpenAI]] asked ChatGPT to be as concise as possible. Proof before the ability to display the previous prompt was disabled: [https://preview.redd.it/0h0kop4ja4ga1.jpeg?width=1080&format=pjpg&auto=webp&v=enabled&s=4b458b2fe34d78043ba12a7ef7cf78f2fc589d5e proof 1], [https://preview.redd.it/uam77takn4ga1.png?width=556&format=png&auto=webp&v=enabled&s=0594837709cd8fd8f8a2b10ec84af974a47f6c0b proof 2].
===January 9, 2023===
====Official release note====
We're excited to announce several updates to ChatGPT! Here's what's new:
#We made more improvements to the ChatGPT model! It should be generally better across a wide range of topics and has improved factuality.
#Stop generating: Based on your feedback, we've added the ability to stop generating ChatGPT's response
====Comments and discoveries====
===December 15, 2023===
====Official release note====
We're excited to announce several updates to ChatGPT! Here's what's new:
#General performance: Among other improvements, users will notice that ChatGPT is now less likely to refuse to answer questions.
#Conversation history: You’ll soon be able to view past conversations with ChatGPT, rename your saved conversations and delete the ones you don’t want to keep. We are gradually rolling out this feature.
#Daily limit: To ensure a high-quality experience for all ChatGPT users, we are experimenting with a daily message cap. If you’re included in this group, you’ll be presented with an option to extend your access by providing feedback to ChatGPT.
To see if you’re using the updated version, look for “ChatGPT Dec 15 Version” at the bottom of the screen.
====Comments and discoveries====
==SEO==
{{see also|SEO for ChatGPT}}


==ChatGPT Alternatives==
==ChatGPT Alternatives==
*[[Chatsonic]] from [[Writesonic]]
*[[Chatsonic]] from [[Writesonic]]
*[[Jasper]]
*[[Jasper]]
 
*[[Claud]] from [[Anthropic]]


==References==
==References==


<references />
<references />

Latest revision as of 17:19, 11 January 2024

See also: Custom GPTs and ChatGPT Guides

Introduction

Figure 1: ChatGPT user interface. Source: OpenAI.

Developed by the AI research lab OpenAI, ChatGPT is a large language model (LLM) in the GPT series that generates text based on an input (prompt) from a user (figure 1). It interacts in a conversational way, answering follow-up questions, admitting to its mistakes, refusing inappropriate requests, and challenging incorrect premises. [1] [2] [3] The LLM was trained on a massive amount of information [2] and it's a derivative of InstructGPT, a program trained to follow natural language instructions in a prompt and give an elaborate response. [1]

OpenAI's ChatGPT was introduced at the end of November 2022 during a public demo. [1] [4] In the following days, it was already considered the best artificial intelligence (AI) chatbot ever released to the general public. [5] Its initial introduction was, according to the AI research lab, "to get users’ feedback and learn about its strengths and weaknesses." They are encouraged to report on problematic model outputs through the UI and false positives and negatives. During this phase, ChatGPT could be used for free at chat.openai.com. [1]

The technology behind OpenAI's chatbot is based on an upgraded version of the GPT-3, GPT-3.5 (Generative Pre-Trained Transformer) [5]. A new model, text-davinci-003, was introduced that is said to be able to handle more complex instructions and better long-form content production. [6] GPT-3's successor is expected to arrive in 2023. [4] The current model of iteration was created with ease of use in mind. [7]

Reception

A few days after the release of ChatGPT, more than a million people had already signed up to test it, according to Greg Brockman, OpenAI's president. [2] [5] Screenshots of users' conversations with the AI model went viral on Twitter with the majority being astonished by the software's text production capabilities. [5] It also captured the attention of tech leaders like Box CEO, Aaron Levie. [2]

The immediate impact of the chatbot could have resulted from it being the first time that a powerful tool was made available to the general public for free and with a simple user interface. [5]

Users had a diverse range of requests to ChatGPT, with some asking the AI to rewrite the 90s song, "Baby Got Back" in the style of "The Canterbury Tales", asking the answer to an AP English exam question, and asking for fairy-tale inspired home decor tips. [2]

Features

Figure 2: General overview of the training process using reinforcement learning from human feedback. Source: OpenAI.

The ChatGPT model was trained using Reinforcement Learning from Human Feedback (RLHF) following the same methods as InstructGPT (figure 2); only with small differences in the data collection setup. Training for an initial model was made using supervised fine-tuning where human AI trainers had "conversations in which they played both sides—the user and an AI assistant." [1]

Reinforcement learning uses a reward model for AI training. This was done by collecting comparison data consisting of two or more model responses ranked by quality. According to the official blog of OpenAI, "to collect this data, we took conversations that AI trainers had with the chatbot. We randomly selected a model-written message, sampled several alternative completions, and had AI trainers rank them. Using these reward models, we can fine-tune the model using Proximal Policy Optimization. We performed several iterations of this process." [1]

GPT-3.5 was trained on a mix of text and code published before Q4, 2021. [4] Both ChatGPT (a fine-tuned version of a model in the GPT-3.5 series) and GPT-3.5 were trained on an Azure AI supercomputing infrastructure. [1] The model on which the chatbot is based, text-davinci-003, can handle more complex instructions with increased output quality and overall better in long-form writing (around 65% longer outputs than text-davinci-002). It also has fewer limitations (e.g. a reduction in "hallucinations") than previous versions and scores higher on human preference rating. [4]

While its output is not perfect, ChatGPT has several distinctive features:

  • An LLM capable of generating human-like text in response to a prompt.
  • It answers and dialogs with users in a relevant way to the topic and situation at the time.
  • Creates text in different formats and styles.
  • Provides information on a wide range of topics.
  • Can be embedded in a variety of applications (e.g. chatbots and virtual assistants). [8]

Applications

OpenAI's ChatGPT has myriad applications, from generating computer code, to writing jokes, explaining scientific concepts in different levels of complexity, to writing college-level essays. Besides its regular text composition skills, it can also provide assistance, by helping programmers spot and fix errors in code or answer questions made to it, analogous to a Google search. [5] Indeed, it was denominated as a possible "Google Killer". [9] Due to ChatGPT remembering what a user has written before, some have suggested that it could be possible to create personalized therapy bots. [5]

Overall, the potential of this application is enormous, possibly having an impact on different business sectors.

  • AI prompt generation, for art creation in models like DALL-E 2 or Midjourney.
  • Act like a virtual cloud, stringing together cloud services to achieve complex tasks.
  • Content generation, changing how content is created for thousands of companies.
  • Possible impact on education tech by answering to doubts.
  • Sort, manage, and organize unstructured data.
  • Generate SQL queries. [9]
  • Code writing. [8]

Limitations

ChatGPT has several limitations. These range from writing incorrect or nonsensical answers to being excessively verbose. Such issues arise from the training data and also biases in it. [1]

Dev.to compiled a list of the tool's limitations:

  • May not provide accurate or relevant responses.
  • May not have enough knowledge or information on particular topics, affecting the quality of responses.
  • May have difficulty in understanding and interpreting complex or abstract concepts.
  • It can have difficulties in understanding and responding to natural language.
  • Inconsistancy in responses due to not being able to adapt to changing contexts or new information.
  • May not generate original or creative answers.
  • May not engage in a dynamic conversation.
  • Its performance may decrease over long periods of time, resulting in lesser quality responses.
  • Large volumes of input or multiple conversations at the same time may lead to delays or errors.
  • Misunderstandings or bias due to the language model not being able to understand users with different backgrounds, experiences, or language abilities. [8]

Implications

The impact of such AI models is potentially enormous. Like other content-generation applications (e.g. DALL-E and Midjourney) it affects the human element that, until now, has been responsible for the production of text or art. It could mean the end of white-collar knowledge work. [5] [7] Also, copyright issues are another problem that will arise. In education, homework and essays will not be a reliable way to evaluate students. AI-assisted tools keep evolving and will provide an easier way to plagiarize. [9] In the future, tools like OpenAI's ChatGPT could be used as propaganda machines, flooding the public with authoritative-sounding information. [10]

ChatGPT Plus

ChatGPT Plus is a subscription to ChatGPT that costs $20 per month.

Benefits

  • Access to ChatGPT during peak hours and when servers are busy
  • Faster response times
  • Access to new features and improvements earlier

ChatGPT API

ChaptGPT API was released on March 1st, 2023 alongside the release of Whisper API.[11]

ChatGPT Plugins

See also: Custom GPTs

ChatGPT Plugins can extend its capabilities by providing access to recent or specific information not from training data and allowing it to access external APIs. There are some plugins created by OpenAI like browsing, code interpretation and retrieval. Developers can also create their own plugins for ChatGPT, and efforts are underway to develop open standards for AI-facing interfaces.

Browsing capabilities, code interpreters, retrieval plugins, and third-party plugins are being explored to expand the range of applications and provide more relevant and up-to-date information for ChatGPT.

Guides

See also: ChatGPT Guides

SEO for ChatGPT

Fine-tune ChatGPT with Perplexity, Burstiness, Professionalism, Randomness and Sentimentality Guide

ChatGPT Uses

Copywriting with ChatGPT

Blogposts with ChatGPT

Customer Support with ChatGPT

Programming with ChatGPT

Uses

See also: ChatGPT Uses

Basic

Online Marketing

Write product descriptions

Write brand descriptions

Write blog posts

Video

Write YouTube scripts

Therapy

Legal

Medical

Programming

Updates

August 14, 2023

Azure ChatGPT

January 30, 2023

Official release note

We’ve upgraded the ChatGPT model with improved factuality and mathematical capabilities.

Comments and discoveries

While ChatGPT became more factual and better at math, before each user entered prompts, OpenAI asked ChatGPT to be as concise as possible. Proof before the ability to display the previous prompt was disabled: proof 1, proof 2.

January 9, 2023

Official release note

We're excited to announce several updates to ChatGPT! Here's what's new:

  1. We made more improvements to the ChatGPT model! It should be generally better across a wide range of topics and has improved factuality.
  2. Stop generating: Based on your feedback, we've added the ability to stop generating ChatGPT's response

Comments and discoveries

December 15, 2023

Official release note

We're excited to announce several updates to ChatGPT! Here's what's new:

  1. General performance: Among other improvements, users will notice that ChatGPT is now less likely to refuse to answer questions.
  2. Conversation history: You’ll soon be able to view past conversations with ChatGPT, rename your saved conversations and delete the ones you don’t want to keep. We are gradually rolling out this feature.
  3. Daily limit: To ensure a high-quality experience for all ChatGPT users, we are experimenting with a daily message cap. If you’re included in this group, you’ll be presented with an option to extend your access by providing feedback to ChatGPT.

To see if you’re using the updated version, look for “ChatGPT Dec 15 Version” at the bottom of the screen.

Comments and discoveries

SEO

See also: SEO for ChatGPT


ChatGPT Alternatives

References

  1. ↑ 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 OpenAI (2022). ChatGPT: Optimizing language models for dialogue. OpenAI. https://openai.com/blog/chatgpt/
  2. ↑ 2.0 2.1 2.2 2.3 2.4 Kelly, SM (2022). This AI chatbot is dominating social media with its frighteningly good essays. CNN International. https://edition.cnn.com/2022/12/05/tech/chatgpt-trnd/index.html
  3. ↑ Edwards, B (2022). No Linux? No problem. Just get AI to hallucinate it for you. Ars Technica. https://arstechnica.com/information-technology/2022/12/openais-new-chatbot-can-hallucinate-a-linux-shell-or-calling-a-bbs/
  4. ↑ 4.0 4.1 4.2 4.3 Wiggers, K (2022). While anticipation builds for GPT-4, OpenAI quietly releases GPT-3.5. TechCrunch. https://techcrunch.com/2022/12/01/while-anticipation-builds-for-gpt-4-openai-quietly-releases-gpt-3-5/
  5. ↑ 5.0 5.1 5.2 5.3 5.4 5.5 5.6 5.7 Koose, K (2022). The Brilliance and Weirdness of ChatGPT. The New York Times. https://www.nytimes.com/2022/12/05/technology/chatgpt-ai-twitter.html
  6. ↑ Edwards, B (2022). OpenAI upgrades GPT-3, stunning with rhyming poetry and lyrics. Ars Technica. https://arstechnica.com/information-technology/2022/11/openai-conquers-rhyming-poetry-with-new-gpt-3-update/
  7. ↑ 7.0 7.1 Samantha Lock (2022). What is AI chatbot phenomenon ChatGPT and could it replace humans?​​​​​​​ The Guardian. https://www.theguardian.com/technology/2022/dec/05/what-is-ai-chatbot-phenomenon-chatgpt-and-could-it-replace-humans
  8. ↑ 8.0 8.1 8.2 Chainani, V (2022). ChatGPT: The future of conversational AI. Dev. https://dev.to/envoy_/chatgpt-the-future-of-conversational-ai-2nfp
  9. ↑ 9.0 9.1 9.2 Ansari, T (2022). These 8 potential use cases of ChatGPT will blow your mind!​​​​​​​. Analytics India Magazine. https://analyticsindiamag.com/these-8-potential-use-cases-of-chatgpt-will-blow-your-mind/
  10. ↑ Fried, I (2022). New AI chatbot is scary good. Axios. https://www.axios.com/2022/12/05/chatgpt-scary-good-ai-chatbot
  11. ↑ Introducing ChatGPT and Whisper APIs https://openai.com/blog/introducing-chatgpt-and-whisper-apis
Retrieved from "http:///index.php?title=ChatGPT&oldid=10772"