make3

5 ms·

1.
by make3 7d ago
honnestly I would be less surprised if a human learned the info and prompted the model in the right direction
2.
by make3 8d ago
Then your choice is either a US mega corp under the Trump partial dictatorship shitshow or a Chinese model distillation factory that could turn on you at any second. I understand that there are users that would prefer another option, especi
3.
by make3 11d ago
Brain dead moral relativism argument. The question is whether eg. a group trying to scam elders out of insurance money or a Columbian cartel to hack local politicians to do blackmail, or South Sudan to hack Darfur or whatever, should be all
4.
by make3 11d ago
Lean is adversarial in a way. Lean is better thought of as a constraint language with a verifier that checks if the constraints are respected, than a programming language. Your job or the LLM's job is to write code that Lean is satisfi
5.
by make3 11d ago
Interesting. Indeed, proving theorems that are stronger and more general "accidentally" than what you really need is not a bad thing.
6.
by make3 11d ago
you would never assume this if you've spoken to any human being, ever
7.
by make3 11d ago
Doing things the hard way is a lot of times "male" bravado (though this is not limited to males in any way ofc), this is a very common thing in junior engineers. Curiosity is good but maturity is trusting that the problem will get
8.
by make3 14d ago
surely you'll get sued by the GLP-1 patent owners like HIMs did if you sell your own product? & can physicians recommend non-FDA approved manufactured drugs?
9.
by make3 14d ago
My assumption is that they're cooking an ultra humongous Gemini 4 Pro release. They certainly have the cash and the compute for it, and it's so obviously the thing to do from a strategic perspective.
10.
by make3 14d ago
That extremely likely just means that they're preparing an omega huge Gemini 4 Pro release and that that's what training right now on most of the compute
11.
by make3 21d ago
brother H-1Bs are transparently vectors for foreign educated workers to come to the US to build lives and eventually get a green card, so the US doesn't have to actually build a functional education system. no one talented will come pa
12.
by make3 21d ago
NVidia has reasonable incentives to keep things open indefinitely, it wants people to use it's GPUs
13.
by make3 21d ago
I think this is good.. They have incentives to keep things free to keep people using their GPUs. I think that's one of the least enshittifying outcomes possible
14.
by make3 22d ago
& the town square forces you to listen to famous gossippers and dumb popular people you've never heard of
15.
RL Contributes to Illegible Reasoning Traces in Hard Problems(arxiv.org)
3 points by make3 22d ago|0 comments
16.
by make3 22d ago
they mention using an open weights model, which one it is doesn't really matter
17.
by make3 22d ago
It will always be wasteful to have a GPU sleeping next to you 99% of the time. Something like OpenRouter is the solution imho, for me at least. I realize that that some people care about privacy more though, and I respect that.
18.
by make3 23d ago
they should unite to block all news from getting to LLMs without LLM providers paying per access
19.
by make3 23d ago
the press release says this explicitly
20.
by make3 23d ago
40M$ to get a marginally better model is surprising, why not just use the free weight models
21.
by make3 23d ago
why chicken or egg? it's way to imagine both evolving together in a mutually reinforcing relationship where the gains in one are direct incentives for the other to continue to evolve, etc
22.
by make3 23d ago
interestingly also, open weight models are also more effectively run in the cloud, so it creates a weird scenario where the frontier labs crash but the compute providers, not as much
23.
by make3 23d ago
well, a stroke of luck until the whole US stock market crashes & everyone's retirement funds get cut 40% I guess when people internalize this. it will have to happen sooner or later though I suppose
24.
by make3 24d ago
There's barely any incentive to publish anything online if the main way to interact with the web is through LLMs that don't click ads or mention your name etc
25.
by make3 24d ago
For me, Opus 5 mostly sucked because of its incomprehensible writing style. Having the system prompt focus on writing in terms that are easier to understand helped a bit
26.
by make3 24d ago
I think it's a prisoners dilemma & that democracy, with its short term incentives, leads to politicians taking the easy way out, even if it creates bad precedent for their countries. obviously democracy is still better than the alt
27.
by make3 24d ago
it's also just waiting for Trump to be less powerful & making sure a bad precedent is not established, which is a good idea
28.
by make3 28d ago
Maybe it's because the United States is not the whole planet
29.
by make3 28d ago
This is my impression as well, that an exaggeratedly precise yet bad at communication with humans way of talking snuck in through reasoning RL, that it might be useful when it talks to itself (CoT)
30.
by make3 28d ago
it started at Opus 4.7, way before watermarking

More ›