
Jim Keller: The Future of Computing, AI, Life, and Consciousness | Lex Fridman Podcast #162
About this video
Lex Fridman interviews chip architect Jim Keller about the future of computing, AI, life and consciousness. Topics include CPU design, branch prediction and why imperfect designs often win.
Why it appears in the archive
Recommended as a stellar long-form interview with someone both very smart and wise, mined for CPU design insights.
AI-generated summary and labels from the source dataset; verify important details against the video. Source & credits.
From the Hacker News discussion
Something I have taken to doing these days is playing some mindless but fun game while listening to long-form interviews or talks. As a random example, these days I'm playing Slipways[1], but Diablo-style games or factory/builder games are also suitable. Meanwhile I like to listen to fantastically insightful interviews with people like Julia Ioffe. You can watch a thousand 1-minute news clips and not gain as much understanding about Putin and the war in Ukraine as from a one-hour interview with her. Similarly, the Telegraph's "Ukraine the latest" is a roughly one-hour podcast I listen to every morning while taking the kid out for a walk. Perun's analysis of the logistics situation also tends to be spot on, and very insightful. I used to listen to interesting characters on Joe Rogan, but Joe turned a bit... right wing nutjob. Lex Fridman is less crazy but also a less skilled interviewer. Still, he's got some stellar interviews like the one with Jim Killer[1]. Do you want to watch a soundbite from a politician in another country talking about something that will never affect you, or do you want to hear from the horse's mouth how your everyday life will change because of new technology? I used to feel I didn't have time for this kind of thing, but overlapping it with gameplay makes it seem like zero time "spent" on it. [1] [VIDEO]
Read original comment ↗K8 seems pretty small after he made the fucking Zen architecture (and Tesla AI). Lex has good interviews with Jim[1], he's not just very smart, he is wise. He started Tenstorrent some time ago with Ljubisa Bajic so I'm assuming this fab is related. Here is youtube channel for Tenstorrent[2], and last time I checked it was getting under 100 views per video... Ian also interviewed him[3] some time ago. Long story short if I had money that would be relevant in this context I would invest really hard into Tenstorrent. Investing time to listen to the guy seems to be not worse. 1a. [LINK] 1b. [VIDEO] 2. [LINK] 3. [LINK]
Read original comment ↗I didn't make the connection until now, but Jim Keller was on the Lex Fridman podcast twice [1, 2] and it was awesome. In terms of founders this is pretty great team. [1] [LINK] [2] [VIDEO]
Read original comment ↗A part about "bad" designs regularly winning both in software and in hardware: [VIDEO] Jim Keller's thoughts of the relationship between productivity and chaos: [VIDEO] His father's later years: [VIDEO] The fast and slow versions of Moore's Law: [VIDEO] On Chris Lattner and Swift: [VIDEO] Do you dream? [VIDEO] Advice to young people: [VIDEO]
Read original comment ↗Per Jim Keller, current source of speed in CPU is better branch prediction and data fetching prediction. It is possible C++ (or some other lang) may have more met information to drive these better, in which case it will be faster than C/assembler. [VIDEO]
Read original comment ↗I heard Jim Keller (the chip-designer) say this (or something similar) recently in a podcast with Lex Fridman. It is at the 24.40 mark in "Jim Keller: The Future of Computing, AI, Life, and Consciousness | Lex Fridman Podcast" [VIDEO] "So there is a graph. Y-axis is productivity. X-axis at 0 is chaos and infinity is complete order. As you improve order, you increase productivity. And at some point productivity peaks, and it goes down again. Too much order -- Nothing can happen... Once you start moving towards order, the force vector that drives you towards order is unstoppable."
Read original comment ↗I'm not qualified to truly understand this, but when Lex first interviewed Jim Keller, Jim basically said regarding processor design -- 'yeah, if you just guess the same result as last time, you get a 50% speed increase.' First Interview (where that poorly paraphrased quote resides): [LINK] Second Interview: [VIDEO]
Read original comment ↗Congratulations on the launch preview! This is exciting, as I've been waiting for this announcement since Jim Keller's callout on Lex Fridman podcast two years ago[0]. I'm curious about the teasing around open source. Obviously the amount of money and the slick product launch dictate a need to capitalize on this pooled expertise. Don't want to give away the game to the hyperscalers. I wonder what the revenue and license models are going to end up looking like. A cloud of their own, professional services to HW manufacturers to optimize their performance, professional services to hyperscalers? Curious if anybody has any ideas beyond the obvious. [0] [VIDEO]
Read original comment ↗