OpenAI’s ‘next-generation’ AI model is behind Microsoft’s new search

Microsoft is making a big AI play with its revamped Bing search engine and Edge web browser, both of which are powered by what appears to be exclusive access to the successor to OpenAI’s popular ChatGPT large language model.

The new AI is unnamed for now, only described in a blog post as following:

…A new, next-generation OpenAI large language model that is more powerful than ChatGPT and customized specifically for search. It takes key learnings and advancements from ChatGPT and GPT-3.5 – and it is even faster, more accurate and more capable.

“Customized specifically for search” was repeated by multiple executives at the Microsoft event, so it must be some agreed-upon language that doesn’t over-commit the model’s capabilities one way or the other. They did call it a “new large language model,” though how much of a step up it is from OpenAI’s previous ones is hard to say. Speculation was that it was GPT-4, but so far that term has not been used. I’ve asked OpenAI for more information and will update if I hear back.

One of the drawbacks of large models like these is the immense computing power required to run them, which has led many a prospective ChatGPT user to wait a few minutes before getting a session going. Microsoft’s focus on scaling makes sense, especially since it will certainly be footing the compute and server bill.

Unlike ChatGPT and the other GPT models, the AI-powered Bing is accessed directly through a normal search interface, and wrapped in a Microsoft-created safety system they call Prometheus. It’s an ominous name to be sure, for although Prometheus was a god of wisdom and cunning, he also famously ended up in endless torture, chained to a rock with a huge bird forever pecking his guts. Let’s hope Bing doesn’t end up the same way.

Image Credits: Microsoft

Prometheus is part of a safety and control layer around the model that acts as a sanitizer and filter, watching for obviously inappropriate or incorrect results. But it also brings in relevant data like location, context, and up to date info to customize or improve the inputs and outputs from the core model.

The next-gen model was also applied to Bing’s search ranking index, “which led to the largest jump in relevance in two decades.” It takes strength to admit that!

We will have more details on the new Bing and the AI model that powers it after today’s event at Microsoft headquarters.

OpenAI’s ‘next-generation’ AI model is behind Microsoft’s new search by Devin Coldewey originally published on TechCrunch