The AI Policy Podcast Scaling Laws, a Chinese AI Ecosystem Update, and a Manhattan Project for AI
Nov 25, 2024
The conversation kicks off with the intriguing idea that scaling laws in AI might be plateauing, raising questions for future investments. A spotlight shines on the competitive rise of China's AI models, particularly with DeepSeek's latest innovation. The urgency for a U.S. Manhattan Project for AI is discussed, emphasizing national security implications. Additionally, the European Union reveals a draft for its General Purpose AI Code of Practice, aimed at guiding responsible development. Exciting updates on AI events and feedback opportunities round out the chat.
AI Snips
Chapters
Transcript
Episode notes
Scaling Laws Might Be Slowing
- Scaling laws describe AI model performance improvement with increased resources like size, data, and compute.
- However, recent reports suggest these laws might be plateauing, impacting future AI investments.
Conflicting Views on Scaling Laws
- Major tech companies are heavily investing based on scaling laws.
- Conflicting views exist, some believe progress will continue, while others observe diminishing returns.
Policy Implications of Scaling Laws
- Policy decisions like permitting reform for data centers reflect a belief in scaling laws.
- The debate's outcome significantly impacts future AI development and policy.