* Theme: LLMs and web-scale RAG with you.com APIs
* Description
Even after countless hackathons and thousands of people working on LLMs all over the world… they still hallucinate, often fail to properly cite sources, and are difficult to keep up to date with the ever changing events of the world. We're bringing together some of the best developers and prompt engineers to leverage databases - external and internal. We’d love to see applications that use the you.com search and RAG APIs. These APIs enable your LLM application to get full access to an index of the entire web for FAST retrieval augmented generation. They have up-to-date news and recent events for much more factual, accurate and relevant LLM outputs. You can find more details on docs.you.com**
Some ideas for projects:
Creative - What LLM app can you build that leverages RAG in creative new ways? We’d love to see examples that combine internal RAG with our external web index. For example, writing a briefing before going into a meeting with a customer. Or combining your own research on a topic with most recent news about it.
Evaluators — evaluating LLMs on rapidly changing information in a scalable way remains an open problem. Leverage academic datasets, your own data gathering skills, GPT-4, or human labelers if you like. The goal is to create a benchmark or use existing benchmarks that show whether whether LLMs are sticking to the facts.
Rules: *All attendees must ship code.* Do not invite other guests. Attendees must have either built an agent or be a very experienced hacker.
11:00am Check in - Doors Open
11:30am Lunch
12:00pm Kick-off and Introduction
12:05 Richard Socher (CEO @ You.com, fmr Chief Scientist @ Salesforce)
12:15pm Secret speakers
12:45pm Lightning Talks (Project proposals from hackers)
1:00pm Hacking Begins!
4:00pm Project Check-In
6:00pm Dinner is Served
8:00pm Demos
* Event sponsor: You.com
* Credit prize co-sponsor: Runpod, Together.AI, MongoDB Atlas Vector Search, etc.