Hacker, Backer, Slacker
New post on ReactionWheel: AI and the Structure of Reasoning https://reactionwheel.net/2023/08/ai-and-the-structure-of-reasoning.html
Great post from @packym@twtr.plus using my starter moat==uncertainty framework to talk about generative AI! https://www.notboring.co/p/when-to-dig-a-moat
It's funny how many people say "but it was adopted faster than any technology in history" as an argument that the tech in question will change the world. Techs that augur a revolution were, in general, adopted exceedingly slowly
I'm looking forward to netflix blocking my five kids from using it wherever they happen to be (college, etc) so I can just cancel and save myself $20/month
I had no idea states did this. "Massachusetts Bars Apple Stock" on the eve of its IPO (NYT, 1980.) MA did not allow IPOs of stocks at prices higher than 25 times earnings. (Apple gained 30% on its first day of trading.)
OK, this one is just ridiculous
"Glamour in the Computer World: Skidmore Graduate a Vice President at Age of 28" (NYT, 1966)
Sarnoff Envisages Social 'Explosion' In Era of Computer (NYT, 1964)
"In a computerized society...executives would cut down on going to offices and wasting time on travel. They would be able to participate in 'high-level' meetings without stirring from home."
"Pocket Computer May Replace Shopping List" (NYT 1962)
Aside from the casual sexism, interesting.
"Electronic communication would tell the store in advance what she needed. She would simply pick up the bundles."
"Taking her computer from her handbag..."
"...the day when a headwaiter could accurately forecast the cocktail a person wanted merely by matching the drinker's characteristics against preferences recorded in his own pocket computer."
"There is no reason to suppose the average girl or boy cannot be master of a personal computer."
What's that saying, we wanted nuclear cars, we got ubiquitous computing? Something like that. (Cover of an IBM pamphlet, 1976)
Interesting that for the hardest problems, AI does better with a massive training set than deducing from assumptions, while humans seem to think they do better reasoning from first principles than using a bigger training set
I am going to train a transformer model on pie recipes, and only pie recipes, so it can generate pecan pies
Back in the 1960s, Robert Noyce recommended incentive pay for one of his top performers at Fairchild Semiconductor. A Fairchild director replied “isn’t the fact that we’re not going to fire them in the morning incentive enough?”
Noyce left soon after to start Intel