Sabitlenmiş Tweet
akhil
25.3K posts
akhil
@fkasummer
For every atom belonging to me as good belongs to you. Technology must multiply human agency.
New York, USA Katılım Ekim 2018
2.2K Takip Edilen8K Takipçiler
@fkasummer Lmao you need measure theory for stochastic integration. It’s the foundation for financial math … I feel like a basic criteria for annoying is that it can’t be useful.
I expected more category theory. Unfortunately, no matter how annoying I find it, I still find it super fun.
English
Interesting that so many people hate measure theory in this thread.
little grey mouse 🐭@mouse_math
what's the most annoying field of mathematics?
English
@fkasummer surprising, given that there is only one right answer (combinatorics)
English
akhil retweetledi
T.S. Eliot on Blake
The Sacred Wood: Essays on Poetry and Criticism
Themes: Blake as isolated genius uprooted from tradition in contrast to Dante.
Kerwin Fjøl@zermatist
@LackOfView This is a similar point to what T.S. Eliot was saying about Blake
English
In frontier research, I can confirm that a rigid routine is counterproductive.
Generally, the pattern I see is that the more originality the work calls for, the more unstructured the time spent on the work.
Zena Hitz@zenahitz
I was a student in four different humanist grad programs (long story.) Only one student of all of those cohorts worked 9-5. It worked great for him. The rest of us worked the way most thinkers work: in fits and starts.
English
Able bodied man super-glues himself to a wheelchair, then complains.
Taelin@VictorTaelin
the current state of Bend2 is: → everything is done → everything works → all tests pass yet I can't launch because the codebase is massive and auditing it is taking forever because each small adjust or bugfix takes a whole day as the AI re-reads everything once again sighs
English
Do people think the cost of inference is the thing preventing a grand slam victory?
taobanker@taobanker
basically what i'm hearing from my guy is that he's hearing from his people that inference costs are gonna work out well so it's really about getting to enough scale to make the training runs economic is this true
English
Halfway in, it's a treat. Making ideas rigorous is easy; you can put some music on and just jam your way through it.
New ideas, on the other hand, need deep concentration and high-quality attention.
akhil@fkasummer
Remarkably, I find myself in a position where I must learn hundreds of pages of pure math over the next 30 hours.
English
@fkasummer like nasm or something? have you seen the work the metareflection lab did?
English
my intuition at this point is this can be built in ~ 100k lines
akhil@fkasummer
one attractive thing about ai native software is that it can stay unorganized longer than traditional software so it unlocks new problem spaces obviously this needs new concepts around runtimes, packaging, etc
English
Pleased with how this thesis developed. Wrestling with ontological & epistemic questions & frontier research is a clear departure from engineering/design mode.
akhil@fkasummer
trying to wrap this new material under traditional methods feels silly–RAGs, structured APIs, etc this material can go from unorganized to organized in a flash. that's insanely valuable. we could have late binding for the software engineering process. possibly witnessing the birth of a new software engineering method
English
People lack an intuition for high complexity. They think high complexity means "too complicated."
Not so. There's absolute difficulty of the kind where you must invent an alphabet, imagine new words, go through the grueling work of finding a grammar, and then solve the problem.
adam@resurrecti0ns
it straight up never occurs to some people that there are some ideas that you can't really understand as a layman because they're too complicated. and it would be a bad thing to pretend otherwise and only pass down warped, extremely simplified versions of those ideas!
English
@babarganesh lol no duels planned and whatever new math emerges from the work will be overshadowed by the empirical results.
English
