Opening Keynote: LLMs, between fantasy and reality - the poin...
summary
The talk examines real-world use of large language models versus hype, citing claims like GitHub Copilot’s 26% boost in pull-request volume and McKinsey’s 20–45% cost savings but emphasizing engineering work needed beyond toy demos. It describes a five-line LlamaIndex RAG demo as useful for prototypes but highlights practical issues—chunking, vector stores, and prompt reformulation—and presents Comodo Health’s production chatbot built from fine-tuned open-source LLMs plus database APIs for exact statistics and UI cards, a project that took over a year to deploy and reduced analyst friction.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 2:49LLM productivity fantasy
- 12:31LLM productivity reality
- 14:18Employment fantasy
- 17:31Product and project fantasy
- 20:25LLM integration in dev
- 27:12Technical challenges reality
- 31:29Innovation and opportunities
related talks
18mBeyond the Chatbot: The Real Impact of LLMs on Observability
51mTowards a Pragmatic Understanding of Large Language Models
49mA Practical Approach to Using LLMs: Peculiarities and Pitfalls
17mArchitecting LLM-Based Chatbots using free open-source LLMs
12mKeynote: How LLMs Have Changed the Way Organizations Do AI
01h 49mAI engineering with open access LLMs that lie, curse, and steal
40mKeynote
54mKeynote: Can you trust your (large language) model?