2023-08-08 15:24:07
Futuristic representation of a brain
GPT-4 performance decreased in three out of four fields tested between March and June, according to the study.
(Foto: Moment/Getty Images)
New York, San Francisco A few weeks ago, the first users reported increasing problems with GPT-4, the most modern language model from the Californian developer of artificial intelligence, OpenAI. Now confirms one scientific study of the elite universities Stanford and Berkeley: The previously leading AI system, which is also used by a number of companies in Germany, has become significantly worse in several areas.
The science team asked OpenAI’s systems in June the same questions they asked in March. If GPT-4 used to be much more reliable, the system has recently been making more and more errors. Since then, the AI scene has been discussing the background.
The finding is a warning for users and companies that are already using GPT-4 or the text robot ChatGPT or are building a business model on them. The most important questions and answers.
KI: What did the scientists test at GPT-4?
Read on now
Get access to this and every other article in the
Web and in our app free of charge for 4 weeks.
Read on now
Get access to this and every other article in the
web and in our app.
Further
#ChatGPT #dumber #Stanford #study