Vanishing Gradients Episode 27: How to Build Terrible AI Systems
31 snips
May 31, 2024 Jason Liu, an independent consultant in recommendation systems, discusses building AI apps, playbook for ML, and avoiding pitfalls. They focus on building terrible AI systems to learn how to prevent failures. The podcast explores consulting in various industries, future tooling, and creating robust AI systems.
AI Snips
Chapters
Transcript
Episode notes
Jason's Journey to LLMs
- Jason Liu studied mathematical physics and was initially dismissive of LLMs.
- He later embraced LLMs after seeing the reasoning capabilities of models like ChatGPT.
Beyond Chat Interfaces
- Chat interfaces limit scalable value delivery in AI.
- Document generation and decision-making offer more leverage.
RAG and Recommender Systems
- Jason's focus on RAG stems from his background in recommender systems.
- He sees strong parallels between recommender systems and RAG applications.