Jim Keller: Moore’s Law, Microprocessors, Abstractions, and First Principles
Lex Fridman Podcast
2020/02/05
Jim Keller: Moore’s Law, Microprocessors, Abstractions, and First Principles
Jim Keller: Moore’s Law, Microprocessors, Abstractions, and First Principles
Lex Fridman Podcast
2020/02/05
In this episode of the Artificial Intelligence podcast, Lex Fridman sits down with Jim Keller, a pioneering microprocessor engineer with a storied career spanning companies like AMD, Apple, Tesla, and Intel. Known for his deep technical insight and innovative thinking, Keller delves into a wide range of topics that bridge computer architecture, artificial intelligence, and the nature of intelligence itself.
The conversation begins with a comparison between computer architecture and the human brain, emphasizing the brain’s distributed memory and complex connectivity. Keller then explores the layered abstractions in computing, parallelism, and the evolving nature of processor design. He discusses the importance of rethinking systems from scratch every few years to avoid stagnation and how Moore’s Law continues to hold due to relentless innovation. The discussion shifts to AI, where Keller examines whether superintelligence could represent a new layer of abstraction, and touches on the philosophical question of whether the universe itself behaves like a computer. He reflects on his time at Tesla and working with Elon Musk, focusing on first-principles thinking and the challenges of building autonomous systems. Toward the end, Keller considers the existential implications of AI, the emotional impact of superintelligence, and concludes with reflections on happiness, the meaning of life, and the endless frontier of human exploration.
07:00
07:00
Modern computers analyze instruction dependency graphs to execute out of order for performance optimization.
17:53
17:53
Modern AI explores noisy algorithms due to noisy input data.
20:47
20:47
The speaker draws an analogy between computer architecture and organizational design.
28:55
28:55
Successful long-term planning often requires redoing systems rather than incremental improvements.
32:57
32:57
Despite decades of predictions about its demise, Moore's Law continues due to countless innovations with diminishing returns that together maintain an exponential trend.
55:51
55:51
AI systems may represent a new stage in human evolution through encoded consciousness
1:00:05
1:00:05
The simulation concept breaks under scrutiny but remains intriguing.
1:03:00
1:03:00
Shrinking transistors has ripple effects beyond just cost.
1:19:41
1:19:41
Factory work is more complex than it appears and can be deeply satisfying.
1:21:07
1:21:07
First-principles thinking strips away assumptions to solve problems from the ground up.
1:28:33
1:28:33
Superintelligence may create interesting possibilities without necessarily inheriting human dark aspects.
1:32:39
1:32:39
The guest prefers the optimism and anxiety of the unknown over reliving the past
