Building With AI Identifying high-value use cases for AI | Tomasz Tunguz (Founder of Theory Ventures)
59 snips
Jun 26, 2024 
Guest Tomasz Tunguz, a VC with AI expertise, delves into AI's evolution, modern data stacks, human evaluations, and high-value use cases. Topics also cover AI disrupting industries, advice for incumbents and startups, successful distribution strategies like Midjourney's on Discord, and the future of AI in non-determinism and professional relationships.
AI Snips
Chapters
Transcript
Episode notes
LLMs: A New Kind of Tech Wave
- LLMs are non-deterministic and chaotic, unlike prior tech waves like mobile or the internet.
- This poses unique challenges for product design and sales due to unpredictable outputs.
LLM Architecture Parallels Modern Data Stack
- LLMs aren't standalone solutions; they require supporting infrastructure like data pipelines and vector databases.
- This mirrors the modern data stack's evolution, but with added organizational and workflow changes.
The Importance of LLM Evaluations
- Evaluating LLM outputs is crucial due to their unpredictable nature and variability across versions.
- Human evaluations, like those used in search engine quality checks, are essential for assessing context-specific outputs.