OpenAI’s new GPT-4 language model can get into top universities without exam preparation
So far, only Plus subscribers can try the improved ChatGPT, but sooner or later all users will have access.
ChatGPT, a popular chatbot from the company Open AI up to now worked on the version of the GPT-3 language model. Just yesterday, March 14, the company announced launch of GPT-4, an improved language model that understands much more context in user requests, and also supports working with images.
According to OpenAI, thanks to GPT-4, a simple chatbot can enter any top universities in Mer without any prior training. According to the company, the GPT-4 performs very well on standard benchmarks. In particular, the language model managed to score very high marks in exams in economics, law, mathematics, physics, biology, chemistry, English, history, etc. In fact, the neural network did better on the exams than most real university students.
OpenAI claims that the new GPT-4 will be able to code multiple programs, generate narrative scripts, answer complex questions with step-by-step instructions, and interact with images.
Access to GPT-4 will appear for ordinary users of ChatGPT gradually. At first, only Plus subscribers will be able to try the improved functionality. Yes, and they, too, the company decided to include all the functionality for now. For example, working with images is currently available only to a number of testers, and not to all Plus subscribers.
“GPT-4 and subsequent language models have the potential to significantly impact society for both good and bad,” according to an OpenAI press release for GPT-4. “We are collaborating with external researchers to improve our understanding and assessment of the potential impact, and to assess the dangerous capabilities that may appear in future systems.”