OpenAI has released its new AI model GPT-4. It is now available today to ChatGPT Plus subscribers. Developers can signup on a waitlist to access the API.
Just like previous GPT models, GPT-4 has also been trained using publicly available data. OpenAI worked with Microsoft to develop a ‘supercomputer’ for training GPT-4.
GPT-4 comes with several improvements over its predecessor GPT-3.5. GPT-4 can accept image and text inputs both and can perform at a ‘human level’ on various professional and academic benchmarks. OpenAI claims that GPT-4 is more reliable, creative, and able to handle much more nuanced instructions than GPT-3.5.
The most notable change is the ability to understand images. It can caption images and even interpret complex images. For the image input capability, OpenAI has collaborated with Bemyeyes.
OpenAI also acknowledges that GPT-4 isn’t perfect It may still make reasoning errors. It can be confidently wrong in its predictions. The model also lacks knowledge of events that occurred after September 2021 and doesn’t learn from its experience. The model could have various biases in its outputs but it has made progress on these.
The company also notes that GPT-4 is 82% less likely to respond to requests for “disallowed” content compared to predecessor.
