– OpenAI has announced the general availability of GPT-4, its latest text-generating model, through its API.
– Existing OpenAI API developers with a history of successful payments can access GPT-4 starting this afternoon.
– GPT-4 can generate text (including code) and accept image and text inputs, an improvement over its predecessor GPT-3.5.
– GPT-4 performs at "human level" on various professional and academic benchmarks.
– GPT-4 was trained using publicly available data, including from public web pages, as well as licensed data.
– It "hallucinates" facts and makes reasoning errors, sometimes with confidence, and fails at hard problems like introducing security vulnerabilities into generated code.
– OpenAI plans to allow developers to fine-tune GPT-4 and GPT-3.5 Turbo with their own data later this year.
– OpenAI is making its DALL-E 2 (image-generating model) and Whisper (speech-to-text model) APIs generally available.
– OpenAI will deprecate older models available through its API to optimize compute capacity, starting January 4, 2024.
– Developers using the old models will need to manually upgrade their integrations by January 4 and fine-tune replacements on the new base GPT-3 models.
– OpenAI aims to provide support to users for a smooth transition and will reach out to developers who have recently used the older models.
– The competition in the generative AI field has grown fiercer since the reveal of GPT-4, with other models expanding their context windows.
– OpenAI is testing the image-understanding capability with a partner, Be My Eyes, before opening it up to a wider customer base.