Microsoft Bing blogs have posted new information about how the new AI-integrated search works, including details of a technology called the Bing Orchestrator that acts as an orchestrator between the Bing site index and the “next-generation GPT” that forms the basis of the new version of Bing.
Microsoft Prometheus Prototype
Microsoft has previously Description of the Prometheus AI modela set of techniques for interacting with the OpenAI model.
The latest blog post provides new details on what Bing’s new AI search is all about.
What stands out first is that OpenAI shared a next-generation version of GPT with Microsoft in the summer of 2022.
This next-generation model features “creative thinking capabilities” that are more powerful than GPT-3.5, the technology that underpins ChatGPT.
It was exposure to this model that inspired Microsoft researchers to create a Prometheus AI model to interact with.
Prometheus is a combination of the Bing search index and what Microsoft calls the next generation of GPT, curated by what is called the Bing Orchestrator.
Grounding by Bing Orchestrator
Among the search and GPT components is the Bing Orchestrator.
What Bing Orchestrator does is receive the search query and orchestrate the process of returning an answer from the Bing index and GPT.
Innovation is how Bing search and answers provide a “foundation” for GPT.
Grounding is a technology that allows AI to use that additional data to provide more context for the answer, particularly context related to what is happening right now or in a geographic area or other contexts.
In general, grounding is a way to provide context for words. Language models understand words in relation to other words, kind of like this is related to that.
Grounding, the way Bing uses it, provides more context.
Bing’s ad states:
“…it provides up-to-date and relevant information for the model, enabling it to answer modern questions and reduce inaccuracies—this method is called grounding.
In other words, the form reasons on data provided by Bing, and is therefore based on Bing data, via Bing Orchestrator.
Thanks to Bing’s foundation technology, Prometheus can also embed citations into sentences in a chat answer so that users can easily click through to those sources and check the information.”
The last part about adding citations to website sources is very important, as it turns the chat answer into a kind of Wikipedia-like answer with links to more information.
These are contextually relevant links, a somewhat unorthodox but sensible way to combine AI with search.
Bing Orchestrator does more, such as creating recursive queries.
This new blog post about AI-enhanced Bing shares:
Prometheus leverages the power of Bing and GPT to recursively generate a set of internal queries through a component called the Bing Orchestrator, and aims to provide an accurate, rich answer to a user’s query within a given conversation context.
Bing doesn’t go into detail about what recursive queries are. Iteration is a reference to the process of repetition.
Presumably, it means iterating a query to produce a larger set of answers and context for GPT.
The illustration visually shows how Bing Orchestrator mediates between the search fragments and the GPT fragments:
An inside look at AI-enhanced Bing
The blog post adds to the growing knowledge of what the future of AI-enabled search will look like.
Many are worried because when they think of AI-based search, what comes to mind is a machine that provides answers and not links to websites.
But this type of model is closer to what is known as artificial general intelligence (AGI).
Artificial general intelligence is a machine that can learn like a human and has comprehensive knowledge.
The state of artificial intelligence is not there yet.
The Bing blog post explains how AI enhances search, just as featured snippets and knowledge graph data enhance search.
Artificial intelligence is already being researched. For example, Google uses artificial intelligence to identify spam (SpamBrain is an artificial intelligence-based spam detection tool).
Another Bing article frames the concept of AI-enhanced search as a co-pilot (Reinventing Search with the New AI-Powered Microsoft Bing and Edge, Your Assistant for the Web).
This includes devising the back-end (internal) as well as the front-end (user-facing) part of the search.
the Research as an assistant article He explains:
“Microsoft Prometheus model. We have developed a proprietary way to work with the OpenAI model that allows us to take better advantage of its power.
We call this set of capabilities and technologies the Prometheus model. This combination gives you more relevant, timely, and targeted results, with improved security.
…Application of artificial intelligence to the basic search algorithm. We also applied an AI model to the core Bing search ranking engine, resulting in the biggest jump in relevance in two decades. With this AI model, even basic search queries are more accurate and more relevant.”
Referring to how implementing GPT in search is more than an improvement to the search experience, Microsoft sees GPT integration as part of the continued evolution of search beyond the classic ten blue links, which can’t provide rich answers to those complex search query queries.
This Microsoft Search Assistant article states:
“New user experience. We’re reimagining how you interact with search, browser, and chat by drawing them into a unified experience. This will open up a whole new way of interacting with the web.”
Read our Bing AI Search explainer article:
Featured image by Shutterstock / Mix and Match Studio