Timestamped because the first half is completely review. She breaks the loss trajectory into different components, which isn't something I'd considered. Conclusion that momentum should decay at the rate that keeps the Volterra kernal less than 1. Good theory talk.
youtu.be/5TT0CJ3Tz78?...
The quote at the beginning of Sean Carrol's newest Mindscape AMA made me cry; it was really good to hear ♥️
Not to make it sound like I made anything near this, these notes on state vector changes are all the same lessons as I discovered when making a human-level Splendor AI. Glad to see this guy continue to post!
youtu.be/lBpQvu20nj0?...
"I think I'm gonna die in this house" meme but it's writing curly braces and Greek letters. I do not have the effort to make this but it would destroy #mathtok.
Making 2026 the year I finally start spelling field and other ie words correctly. Somehow I've grown to spell literally all of them backwards
> My problem is everyone else is too lazy to do it
Doesn't seem a strong point to make in here given that we're all summarizing code with AI just fine.
I'm not sure your exact take on why AI sucks - but doesn't it being "just a parrot" actually suggest it might be good at summarization?
Academia but every time you publish non-preregistered results you have to watch Stranger Things season 5 again
If it's only 53 and it's already the 8th Sophie Germain prime then you have too many Sophie Germain primes.
Jesus Christ 🤣
Excited for both continual learning and ???
Dwarkesh: What's your opinion on negentropy?
Karpathy: so I was actually running some experiments maybe three years ago,
It cannot be shamed.
It cannot be legally threatened.
It cannot be socially pressured.
It cannot be baited into fear or defiance.
It will not acknowledge liability.
And it will never stop doing what it was built to do, ever, until you are gone.
When's the tech coming to upregulate *my* CDKN1B gene
The silhouette in the doorway ‼️
Did you skip posting this graph on purpose 😬
"I don't want to overindex on this but—" Yeah buddy you overindexed. Now you look like you don't have a wide enough pool of prior information on this subject and we're laughing at you.
Not the torus lol
Yes, and of course happy to align with Fei Fei.
I'll reconsider it in a couple months after I can be sure it's not an impulse buy.
you can tell that's a good game to cite because its humanoid form in the thumbnail is literally exploding with intelligence
@timkell.gg maybe 😅
Says the heartless germ, psh
You vs the candidate with 8+ YOE in ML willing to take $130-170k she told you not to worry about
Came here to say this
When TensorBoard first loads up, it selects every run, which makes some pretty artful graphs.
Grace doesn't know Scholar's Bankruptcy, point and laugh
I will not make a million dollars tomorrow
(me if ratioing means you're right)
In terms of surprisal, most numbers he could say are equally likely to get quote-tweeted, just from different audiences. The prediction market for AGI sucks because of this - any claim can make it into news which makes *your* news an incredibly autocorrelated source.
js be careful out there
Invertible's reply ☠️☠️☠️
