

Chips and the Future of Computing
10 snips Feb 13, 2025
Mark Horowitz, a leading semiconductor expert and Chair of Electrical Engineering at Stanford, joins Sebastian Elbaum, a technologist focused on AI implications, to discuss the future of computing. They dive into the evolving landscape of chip manufacturing, the challenges posed by diminishing returns of Moore's Law, and how hardware strides are essential for AI advancement. Their conversation navigates the geopolitics of semiconductor exports, the need for a skilled workforce, and the exciting, yet cautious, outlook on quantum computing.
AI Snips
Chapters
Transcript
Episode notes
Moore's Law Evolution
- Moore's Law, stating that transistor density doubles periodically, has held for a long time, driving down computing costs.
- However, while transistor density continues to increase, the cost-per-transistor decrease has stalled, impacting cost scaling.
Shift to Software and Optimization
- The decreasing cost of computing has driven a shift towards software solutions, making features cheaper to implement.
- As cost scaling slows, optimizing the entire computing stack becomes crucial, as we can no longer rely on hardware getting cheaper.
DeepSeek's Cost-Effective AI
- Chinese startup DeepSeek developed AI models with performance comparable to US companies, but at significantly lower training costs.
- This was achieved due to constraints on access to high-end hardware, forcing them to optimize their processes.