Checkpoint California

Sean Forbes & Charles McGuinness

We're Living in the Twenty-Thirties — Charles McGuinness on AI

OCT 27, 202644 MIN

Description

Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode! Sean turns the microphone on his co-host. Charles McGuinness, who took a master's in AI in 2019 and has built with it every day since 2023, says everything the boosters and the skeptics claim is true at once — and explains why. How AI got here a decade early: not better chips but doubled money, over and over, like the Space Race. Why that is ending, told through the Canada lynx and the snowshoe hare: the transformers, memory chips, power and water have run out, and so has the patience of towns asked to host data centers. The Luddites, who weren't against machines but against shoddy ones. What a model actually stores, why it can't recite Pride and Prejudice, and why asking it who was Cuba's first president after the revolution is the wrong job for it. Who really loses their job, and why "we replaced them with AI" is the dog-ate-my-homework excuse for layoffs the tariffs caused. Then the California answer: health care, basic income, childcare and school lunches, so that people are safe enough to take the risks AI makes possible — including, Charles says, this podcast. In this episode 00:22 — Charles, the show's other host 01:18 — A master's in AI, by dumb luck 03:06 — Is it the next dot-com bubble? "Everything you said is true" 03:57 — Moore's law, and buying the future with money 06:08 — Grandparents: horses to the space shuttle 07:07 — The Space Race, and why we stopped going to the moon 08:46 — The Canada lynx and the snowshoe hare 10:20 — No transformers, no memory chips, no patience 10:58 — The Long Beach data center that became an industrial park 12:28 — California makes data centers pay their own way 13:40 — Tropical storms, El Niño and Los Alamitos High 14:16 — The Industrial Revolution and its Luddites 18:20 — A table saw, not a brain 19:10 — Whose electricity, whose blueprints: the training question 19:32 — What a model holds, and why it can't recite Pride and Prejudice 21:24 — The first president of Cuba, and hallucination 23:46 — "Probably helium": what AI doesn't know about certainty 25:12 — The only driver of the economy 25:55 — Which jobs go: paralegals, leasing agents, maintenance crews 28:02 — The CEO's two speeches: the dog ate my homework 30:09 — Robots, cars and China 31:32 — Amazon, the contractors and the bottles 33:08 — The Star Trek future, and universal basic income 34:11 — Health care first, then Friedman's negative income tax 36:03 — Childcare, school lunches, and feeling safe enough to take risks 37:30 — How this podcast is made 39:40 — Too long, didn't listen 40:59 — Experts in government: "you don't want me in charge" 43:03 — Sign-off, and the Discord Facts, sources and further reading. We checked what was said in this episode, right and wrong, with links, and added the reading list Charles prepared the show from: checkpointcalifornia.com/009