ChatGPT, a popular chatbot from Openai to date, worked on the version of the GPT-3 language model. Just yesterday, March 14, the company announced the launch of GPT-4-an improved language model, understanding much more context in user queries, and also supporting work with images.
According to Openai, thanks to the GPT-4, a simple chatbot can enter any top universities a measure without any preliminary preparation. According to the company, the GPT-4 copes with standard tests very well. In particular, the language model managed to score very high points for exams in economics, jurisprudence, mathematics, physics, biology, chemistry, English, history, etc. In fact, the neural network coped with exams better than most real students of higher educational institutions.
Openai claims that the new GPT-4 will be able to write a code for several programming, generate narrative scenarios, answer complex questions using step-by-step instructions and interact with images.
GPT-4 access will appear in ordinary ChatGPT users gradually. At first, only PLUS subscribers will be able to try improved functionality. Yes, and the company also decided to turn on all the functionality so far. For example, work with images is available so far only by a number of testers, and not to all PLUS subscribers.
“GPT-4 and subsequent language models can significantly affect society both for the good and to the detriment,” the Openai press release for the GPT-4 said. “We cooperate with external researchers to improve our understanding and assessment of potential exposure, as well as assess the dangerous opportunities that may appear in future systems.”