Recently, Microsoft announced the launch of a new AI-enabled version of its search engine, Bing, which was built on top of OpenAI’s GPT models. However, at the time, neither Microsoft nor OpenAI confirmed which version of GPT was being used, except to say that it was a next-generation version of the model that powers ChatGPT. Today, OpenAI has announced the release of GPT-4, a significant update to its previous GPT-3.5 model. As it turns out, Microsoft’s new Bing search engine was already using the new GPT-4 model all along.
In a recent announcement, Microsoft confirmed that the new Bing search engine is powered by a customized version of GPT-4 that has been optimized specifically for search purposes. According to Yusuf Mehdi, Microsoft’s Corporate Vice President and Consumer Chief Marketing Officer, if you have used the new Bing preview in the last five weeks, you have already experienced an early version of this powerful model.
It is interesting to note that Microsoft was confident enough in the new GPT-4 model to stake its reputation on it and invest in the more complex model. However, Microsoft is also using its own Prometheus model in combination with GPT-4 to provide more up-to-date information and to put guardrails around OpenAI’s model.
After a rough start, during which the new Bing was prone to hallucinations, Microsoft quickly iterated on the search engine in recent weeks. The company placed a number of restrictions on it early on but has since loosened up again. Microsoft recently extended the number of possible turns in a conversation to 15 and now allows users up to 150 chats per day.
Despite all of this, Bing still insists that it does not run on GPT-4. However, it seems someone needs to tell Sydney that the information is no longer under NDA. In any case, if you want to give the new GPT-4 model a try, you can head over to Bing or join the waitlist if you haven’t already done so.