Expanding the Capabilities of LLMs Through Community Collaboration
Talkat Open Source Summit Europe 2023 · on Thursday, 2023-09-21· at 11:55· Room 0D-2-0D-3 (Floor 0), Abandoibarra Etorb., 4, 48011 Bilbao, Bizkaia, Spain·
summary
Community open‑source tools show concrete upgrades: embeddings plus an indexer (LLama Index) let a 14,000‑character web page be handled by chunking into 18 semantic requests so GPT‑3.5 can summarize it, LangChain provides agent orchestration and AutoGPT patterns let LLMs generate and run code, use file search and shell tools, and even run a Python "roll a die" action that returns 3. The claim is that embeddings + indexing overcome token limits, 16k token models lose task quality, and agents execute Python and shell reliably.
AI-written summary · about
chapters (25)
- 0:00Introduction
- 0:29Who are you
- 1:17Agenda
- 3:05Breakthrough
- 3:39Research Papers
- 4:06Risks
- 5:40Models
- 6:11Ecosystem
- 7:35Context Window
- 9:05Embedding
- 10:58San Antonio Index
- 11:52Document summarizer
- 12:27Index
- 13:10Summary
- 13:41Knowledge and Capabilities
- 15:20Agents
- 16:33Agents Evolution
- 18:10Transfer
- 18:58Activity
- 19:48Spoilers
- 28:03Launch
- 29:09Other Useful Tools
- 30:16Test Driven Engineering
- 31:28Conclusion
- 32:35Feedback
related talks
37mExpanding the Capabilities of LLMs Through Community Collaboration
27mProductionizing Open Source LLMs
01h 18mThe Enigma of LLMs: on Creativity, Compositionality, Pluralism, and Paradoxes
57mLLMs gone wild
24mHarnessing the Power of LLMs in Advanced Analytics
36mLLMs and AI agents demystified
36mBuilding & Deploying LLM Apps
23mPanel