

Episode 01: Kelvin Guu, Google AI, on language models & overlooked research problems
Dec 15, 2020
AI Snips
Chapters
Transcript
Episode notes
From Biology to AI
- Kelvin Guu's biology research involved tedious tasks like counting cells and monitoring rat behavior.
- This sparked his interest in automation and ultimately led him to AI research.
Knowledge Base Limitations
- Knowledge bases are effective for representing and reasoning over information, but struggle with coverage.
- Real-world applications often require world knowledge that isn't captured in structured schemas.
REALM and World Knowledge
- REALM uses retrieval to augment language models with world knowledge from text documents.
- This allows updating knowledge without retraining the entire model.