What Do We Know About GPT-4 So Far?
OpenAI’s GPT-4, the new large language model, is the most secretive release from the company yet, despite its long-awaited unveiling. OpenAI hasn’t revealed why the model is bigger and better than its predecessor, ChatGPT. OpenAI’s chief scientist, Ilya Sutskever, mentioned that they couldn’t comment on this matter because of the competitive nature of the industry.
GPT-4 can respond to both text and images, making it a multimodal large language model. It has been designed to understand the world better, which could tackle the weak points of language models such as spatial reasoning. Although OpenAI claims that GPT-4 performs better than ChatGPT, there is no information available on how large the model is, how it was built, or its training techniques.
Access to GPT-4 will be available to users who sign up to the waitlist and for subscribers of the premium paid-for ChatGPT Plus in a limited, text-only capacity. OpenAI claims that GPT-4 is 82% less likely than GPT-3.5 to respond to requests for content that OpenAI does not allow, and 60% less likely to make up responses. OpenAI used the reinforcement learning method to improve GPT-4, similar to what was used for ChatGPT. The team used the model to generate inputs that led to biased, inaccurate, or offensive responses and fixed the model to refuse such inputs in the future.