Eschatology
Palmer Luckey says he's never had an original idea.
Here's why he looks to 70-year-old sci-fi to build the future:
"One of the things that I've realized in my career is that nothing I ever come up with will be new. I've literally never come up with an idea that a science fiction author has not come up with before at some point, which makes sense."
"They don't have to make things, and they don't have to wait till the right moment. You remember I started Oculus at just the right moment for it to succeed."
"Some of the stuff that I'm building today, for example, in the AR/VR space around augmenting the vision of soldiers, these are ideas that are from 1959 like Starship Troopers novels, right? Like these are old ideas that have only very recently become technologically feasible."
"The idea of autonomous fighter jets, that's been around for about a hundred years."
@PalmerLuckey
7
How to prevent doom scrolling
Scroll without watching till you see the least amount of likes
Say 1000s to sub 500 for example
Scroll up and watch till you touch the first post
That's your scroll per app open or for the day
9
At some point the current AI will be so far ahead that using models from 5 years ago , will be like having a small god in your home
This is f*cking insane
Everyone is arguing about whether recursive self-improvement has started
An ex-Anthropic researcher says an IPO roadshow slide was a big factor in his decision to quit. Dwarkesh says it's already here. Others call it hype
Jensen Huang gave the most practical answer I've heard, to Ezra Klein:
"I think that RSI is fundamentally how things are done"
"We use software to design a computer to run software to design a computer"
In 4 minutes he walks through how the loop actually works:
→ an agent runs a task and studies which path worked
→ it writes that down as a skill and a memory
→ next time it's faster
→ that data trains the next model
"What used to take a year to pre-train something now takes several hours"
Then the part nobody quotes:
"Don't ship Nvidia any products that humans did not in the loop evaluate. Please don't do that"
His own company runs the other way round from the labs:
"10%, 20% of our company is dedicated to design. 80% is dedicated to verification"
Most labs, in his words, are 80% capability and 20% safety and evals
Old way: race on capability, test at the end
New way: the loop speeds up, so verification gets the compute
"AI needs to accelerate to be safe"
Is RSI already here, or is it just engineering with a scarier name?
7
hold your breath count 1-10 retweeted
Rare esoteric wisdom that can only be gained through suffering.







