Sabitlenmiş Tweet
Cam Howe
664 posts
Cam Howe
@camhowe1729
a humble rebel interested in basement agi, higher order thinking and positive sum games.
The Milky Way Katılım Nisan 2024
2.1K Takip Edilen117 Takipçiler
Cam Howe retweetledi
@DearS_o_n Pharma. Food industry. Schools. Banking system. Marriage. Taxation. Luxury. “Independent” media. “Non-governmental” organizations. Cities. Countries. Reality itself.
English
@a1zhang @askalphaxiv @pangramlabs the real funny thing will be when it shows the llm outputs as human text and other parts as ai-written.
(i stay unconvinced by this ai slop detection thing)
English
@askalphaxiv @pangramlabs Worth mentioning that it seems to label LLM output examples (e.g. in the appendix for trajectories) as part of the % of the paper that's AI-generated, which definitely shouldn't be marked
English
Ever suspected a paper you’re reading is AI slop?
You can now turn on AI detection mode on alphaXiv to visualize what is written by an AI and what is not.
Now available for every research paper indexed on arXiv.
Integrated with the latest detection model from @pangramlabs 🚀
English
@BreatheLesss this can also be said for anything that one believes is the highest virtue. the "moral validation" part could be the bigger problem.
English
Cam Howe retweetledi
The "comfort zone" is arbitrary. It is very self-helpey slop that basically translates to "you should do shit you don't like and feel bad every day"
I think it's more useful to think of it in the sense of setting goals or doing things that you aren't 100% sure you're going to succeed in.
English
Cam Howe retweetledi
the power of billions of least significant bits outweighs that of a few most significant bits
English
i'm a big admirer of the stripe press though.
Cam Howe@camhowe1729
i think it is a stupid chart. it is like boasting about learning relativity in a month while einstein tooks years to invent it.
English
i think it is a stupid chart. it is like boasting about learning relativity in a month while einstein tooks years to invent it.
Patrick Collison@patrickc
Congratulations to Alex and the whole team at MSL. As a sucker for all things speedy (patrickcollison.com/fast), I thought this was an impressive chart:
English
Releasing FLUX.2 Small Decoder: a faster, drop-in replacement for our standard decoder.
→ ~1.4x faster
→ Lower peak VRAM - decode larger images without running out of memory
→ Minimal quality loss
→ Works with FLUX.2 out of the box
Especially impactful for real-time and larger resolutions pipelines.
English
@kalomaze what do you make of the FER paper? it says evo search has much better representations than GD.
arxiv.org/abs/2505.11581
English
and i think this is a good thing, actually!
evo search, genetic algos, things of that nature i feel mostly have a role in narrow cases where backprop is incoherent (think traveling salesman esque stuff, simulated annealing, black box optimization) but otherwise lack this property
English
probably because zeroth order optimization gives you questionable operational guarantees about the structure of the learned solution; backprop (with residuals) imposes structure over the solution space that favors fairly composable internal representations
François Chollet@fchollet
One thing about DL researchers that has always been surprising to me, is that a lot of them have never been exposed to forms of learning other than fitting the parameters of a curve via gradient descent, and are even unable to conceive that there might exist other options
English
the best thing about japanese films is their idea of celebrating the ordinary.
(this art is based on a scene from the movie "linda linda linda", which is about a punk rock girl band).
동동@dongDD_012
린다 린다 린다 中
English
i always like to think of it like this:
real continual learning : current continual learning :: antifragile : robust
Kenneth Stanley@kenneth0stanley
The more you learn the easier it should be to learn more. The key word is easier. What could be more natural? That’s the real puzzle of continual learning. Merely avoiding brain damage from accumulating additional knowledge is barely scratching the surface.
English
The beauty of markdown is that it is useful. But it hides secondary lesson: experts are not always the best people to solve problems. Anyone with a strong CS background would never have designed markdown, because it cannot be parsed cleanly as a context-free grammar.
English