{"id":1108638,"date":"2024-12-06T18:40:37","date_gmt":"2024-12-07T02:40:37","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?post_type=msr-blog-post&p=1108638"},"modified":"2024-12-07T15:36:15","modified_gmt":"2024-12-07T23:36:15","slug":"microsoft-research-neurips-beta-ai-experience","status":"publish","type":"msr-blog-post","link":"https:\/\/www.microsoft.com\/en-us\/research\/articles\/microsoft-research-neurips-beta-ai-experience\/","title":{"rendered":"Microsoft Research @ NeurIPs beta AI assistant"},"content":{"rendered":"\n

To help visitors to the Microsoft Research website get a sense of the research presented at NeurIPs 2024 by Microsoft research teams, we have created a beta AI assistant to accompany our standard event web page<\/a>. This AI assistant was built using Azure Open AI and GPT 4.0 and designed to develop a broad understanding of the work presented in the more than 100 research papers, identify key trends in the research directions and methods used and group the work by research topic area. <\/p>\n\n\n\n

The goal of the AI experience was to try to go beyond a typical chat experience, which can put the burden on the user to articulate in detail what they want to find or learn more about. We also wanted to try to break through the “silos” of information that a typical list of papers creates. If a visitor is interested in reading papers a typical list of papers with descriptions and faceted search can work well. But if you don’t want to read 100 papers and quickly want to know what some of the more interesting ways teams are using small language models, that presents a different challenge. <\/p>\n\n\n\n

To help aid in that kind of topic\/insights scenario, we build a data ingestion pipeline that uses an LLM to review each paper and extract key data points and insights, based on a “data recipe” given to the LLM. From that prompt, the LLM develops an ontology to categorize the work and for those various research topics, generates summaries and insights about the work within each of these topic areas. That data can be shared directly with the user via the application UX and provides more powerful and effective grounding data for the AI chat experience. <\/p>\n\n\n\n

As you browse through the page, you’ll see a mix of typical web site ux as well as chat interfaces and cues for suggested prompts you’d expect from an AI chat. Merging these experiences together presented some interesting challenges, and like all good prototypes, helped teach the team a lot about what works and what needs more thinking. We hope that exposing these research topics and themes helps better orient the reader to the content and makes it easier to explore the broad corpus of data to find the things that interest them the most.<\/p>\n\n\n\n

We hope that these changes help provide a more reliable experience for people visiting the page, and provide better wayfinding and orientation that lets you better understand the trends and emerging topics that the latest research from Microsoft is focused on.<\/p>\n\n\n\n

This experience is very much an experiment. The text is all AI generated so may be somewhat inconsistent, and the user experience is something new, trying to marry the typical AI chat with more traditional web navigation and UX elements. More experiments are ahead, so we value any feedback or suggestions you might have<\/a>.<\/p>\n\n\n\n

<\/p>\n\n\n\n

Thank you,<\/p>\n\n\n\n

The Microsoft Research Web team<\/p>\n","protected":false},"excerpt":{"rendered":"

To help visitors to the Microsoft Research website get a sense of the research presented at NeurIPs 2024 by Microsoft research teams, we have created a beta AI assistant to accompany our standard event web page. This AI assistant was built using Azure Open AI and GPT 4.0 and designed to develop a broad understanding […]<\/p>\n","protected":false},"author":35981,"featured_media":0,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":true,"_classifai_error":"","msr-content-parent":0,"msr_hide_image_in_river":null,"footnotes":""},"research-area":[13556],"msr-locale":[268875],"msr-post-option":[269151],"class_list":["post-1108638","msr-blog-post","type-msr-blog-post","status-publish","hentry","msr-research-area-artificial-intelligence","msr-locale-en_us","msr-post-option-reject-for-river"],"msr_assoc_parent":[],"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-blog-post\/1108638","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-blog-post"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-blog-post"}],"author":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/users\/35981"}],"version-history":[{"count":3,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-blog-post\/1108638\/revisions"}],"predecessor-version":[{"id":1110489,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-blog-post\/1108638\/revisions\/1110489"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=1108638"}],"wp:term":[{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=1108638"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=1108638"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=1108638"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}