When Microsoft declared the new AI-enhanced Bing, which it constructed upon OpenAI’s GPT models, neither firm would demonstrate which edition of GPT was being utilized beyond expressing that it was a next-generation variant of the model that powered ChatGPT. Recently, OpenAI declared GPT-4, a noteworthy improvement to GPT-3.5. As it appears, Bing was utilizing it all the time.
Yusuf Mehdi, Microsoft’s corporate VP and consumer chief marketing officer, mentioned in today’s announcement that the new Bing works on GPT-4 that has been modified for search functions. He also added that if anyone has used the new Bing in the last five weeks, they would have already experienced the effects of this powerful model.
When Microsoft released the new Bing, there had been many rumors that it used GPT-4, so it is not shocking, but it is fascinating to see that Microsoft felt secure enough in the model to use it and pay for the more intricate model. Microsoft is using both GPT-4 and its personal Prometheus model to give fresher information and control OpenAI’s model.
Microsoft’s launch of Bing didn’t go as smoothly as they had hoped, as the new search engine had a few issues. They quickly fixed the problems and set some restrictions, but now they are loosening the rules. Just yesterday, they allowed conversations to go up to 15 turns and they are now allowing users to have up to 150 chats daily.
If you would like to experiment with the new GPT-4 model, go to Bing or join the waiting list if you have not already done so.
Bing insists that it does not use GPT-4, even though Sydney might think it is still under a nondisclosure agreement.