OpenAI, the abundantly funded startup responsible for ChatGPT and DALL-E, has released the next generation of its language model called GPT 4 intended to be more sophisticated and accurate than previous iterations.
Large language models are massive troves of data that power applications like ChatGPT and Microsoft's Bing chatbot.
"GPT-4 is more creative and collaborative than ever before," OpenAI said on its website announcing GPT4. "It can generate, edit, and iterate with users on creative and technical writing tasks, such as composing songs, writing screenplays, or learning a user’s writing style."
Developers who want access to the GPT4 API will have to sign up on a waitlist. The general public can use GPT4 by signing up for ChatGPT Pro for $20 a month.
"We spent 6 months making GPT-4 safer and more aligned," the company said. "GPT-4 is 82% less likely to respond to requests for disallowed content and 40% more likely to produce factual responses than GPT-3.5 on our internal evaluations."
OpenAI also made public a number of partnerships with companies that are building applications on GPT4, including Stripe, Duolingo and Morgan Stanley.
However OpenAI warns that the underlying problems with previous GPTs still exist such as hallucinations (when the AI makes something up) and social biases.