BG2Pod with Brad Gerstner and Bill Gurley Ep20. AI Scaling Laws, DOGE, FSD 13, Trump Markets | BG2 w/ Bill Gurley & Brad Gerstner
178 snips
Nov 21, 2024 The discussion dives into AI Scaling Laws and their impact on tech giants. Insights on the future of inference and the importance of data quality unfold. They also explore advancements in AI memory and actions. A detailed look at National Full Self-Driving Regulations reveals challenges ahead for autonomous vehicles. Comparisons between Tesla and Waymo highlight market disruption in the auto industry. Finally, the chatter shifts to restructuring government departments and the need for increased efficiency, especially in light of national debt.
AI Snips
Chapters
Transcript
Episode notes
Scaling Laws Plateau?
- AI scaling laws, especially in LLMs, might be reaching a plateau.
- This shift could impact expectations and resource allocation in the field.
Diverse Scaling Vectors
- Pre-training is just one aspect of AI scaling; other vectors include post-training, inference, and data quality.
- Focusing on diverse scaling methods might yield better overall improvement.
Shift to Post-Training
- A potential slowdown in pre-training gains could shift focus to post-training and data quality.
- This emphasizes data flywheels and architectural breakthroughs for differentiation.