Mar 14, 2023 - Technology

OpenAI releases major new chatbot engine

headshot
Illustration of a robot emoji inside quotation marks.

Illustration: Maura Losch/Axios

OpenAI on Tuesday released GPT-4, a more powerful update of its text generator engine that will power its own AI products as well as those from Microsoft and other companies.

Why it matters: The world is already enamored with ChatGPT, which is based on an earlier version. Improvements could go a long way to addressing criticisms of the hot new chatbot-powering tool about its errors and other shortcomings.

Details: OpenAI says GPT-4 "surpasses ChatGPT in its advanced reasoning capabilities." For example, it says that whereas ChatGPT could score in the 10th percentile on the bar exam taken by lawyers, OpenAI says GPT-4 can score in the 90th percentile.

  • Not everyone will be able to access GPT-4 immediately, however. OpenAI said it will be available in a limited capacity to paid ChatGPT Plus subscribers via chat.openai.com and there is a waitlist for businesses and developers looking to incorporate GPT-4 via an API.
  • Early customers include Duolingo, Morgan Stanley, Khan Academy and the government of Iceland. Microsoft also confirmed that GPT-4 is what is powering the new Bing.

The new version can accept and generate longer entries — up to 25,000 words. It can also start generating captions and other information with an image as a starting point.

  • OpenAI said it incorporated human feedback in the testing to make GPT-4 both safer and more natural than earlier versions of the technology.
  • The company says GPT-4 is 82% less likely to respond when asked for content it doesn't allow and 40% more likely to produce factual responses than GPT-3.5 in internal testing.

Yes, but: OpenAI acknowledged even the new engine has its limitations, including incorporating societal bias and the propensity to "hallucinate," or make up convincing sounding but false information. And, like its predecessor, some of the new model's safety guardrails can be overcome by targeted efforts.

  • "While we have significantly improved GPT-4’s ability to refuse some tasks that have the potential to be harmful, GPT-4 still has known limitations that we are working to address," the company said on its website.

The big picture: Generative AI has taken the tech world by storm since the November release of ChatGPT, with companies large and small rushing to add some form of the new technology into their products.

  • Google earlier today outlined how it will use generative AI to help businesses. It will offer tools that companies can use to allow generative AI engines to scour through corporate data. AI tools being added to Workspace will help summarize e-mail, craft marketing campaigns and rewrite documents.
  • Google's new features are currently only available to a select group of testers.
  • Microsoft has scheduled an event for Thursday to talk about how it will build generative AI into its business products, including Office apps, such as Word, Excel, PowerPoint and Outlook.
  • Anthropic, an OpenAI rival, today formally announced Claude, its chatbot which is being used by a range of companies including DuckDuckGo, Notion and Quora, among others.

Go deeper: How ChatGPT became the next big thing

Go deeper