aoeusnth1

11 ms·

91.
by aoeusnth1 1y ago
Not feasible.
92.
by aoeusnth1 1y ago
You don’t own your marriage either? What exactly is the distinction you’re trying to make here, that you can hang onto your marriage even if your spouse doesn’t want it?
93.
by aoeusnth1 1y ago
You used the word reinforcing, and then asserted there's no reward function. Can you explain how it's possible to perform RL without a reward function, and how the LLM training process maps to that?
94.
by aoeusnth1 1y ago
Should companies never pay dividends?
95.
by aoeusnth1 1y ago
If training is cheaper, then costs go down and models become both cheaper and more profitable. Training be scaled up to make the models smarter and larger. Distillation can bring those benefits to smaller models. Training is a pure cost - i
96.
by aoeusnth1 1y ago
Statistically, senior engineers are less likely to accept AI suggestions compared to juniors. This is just a tidbit of supporting evidence for the suggestion that juniors are not properly reading and criticizing the ai output.
97.
by aoeusnth1 1y ago
All of the top upvoted comments are calling or implying this is a bubble. It seems to me the majority loves to imagine themselves persecuted in their home turf.
98.
by aoeusnth1 1y ago
That's not a stable equilibrium if the hits have a large enough effect with respect to the movement of the balls. The internal circle will create disturbances against both sides of the inner circle, but the outer ball will have to trav
99.
by aoeusnth1 1y ago
Wow, look at the crowd of NN doubters in the comments there. I see the quality of foresight in the commentariat hasn’t improved given the state of this thread, either.
100.
by aoeusnth1 1y ago
Does a 20% decrease result in a -50% discount? Why would it be nonlinear?
101.
by aoeusnth1 1y ago
Yes, Google uses trunk based development for almost everything. Even Android is now on trunk based development (previously a branchy mess) and heavily uses sets of feature flags for promotion / demotion through quality lifecycles. I’m
102.
by aoeusnth1 1y ago
I don’t understand why PRs are being discussed here at all – doesn’t anyone serious use trunk based development with code reviews? How is that different from a PR? Seems like we need to disentangle the discussion of whether code review is g
103.
by aoeusnth1 1y ago
When forced to point to a concrete issue by tooling, I often just end up rewording it - "Ideally this should X, but it Y". Comments are fine, lower overhead and don't require re-triaging later, and have all the context to imm
104.
by aoeusnth1 1y ago
Yet, richer people have fewer children. It isn't clear that reducing costs (making people richer) would actually increase birth rates
105.
by aoeusnth1 1y ago
Probably because lichess.org exists?
106.
by aoeusnth1 1y ago
The standard model is computable, so no. Physical law does not allow for non-computable behavior.
107.
by aoeusnth1 1y ago
Can you explain in what way the incentives are misaligned here? Or did that mis- prefix slip in there because of how you feel overall about the law?
108.
by aoeusnth1 1y ago
Are you intentionally sandbagging the LLMs to prove a point, or do you really think 4o-mini is good enough for programming? Even 2.5 flash easily gets this https://imgur.com/a/OfW30eL
109.
by aoeusnth1 1y ago
No, it's always strictly worse because you could have bought bonds or deployed the capital in some other way with that money.
110.
by aoeusnth1 1y ago
Or spend on time building test selection systems…
111.
by aoeusnth1 1y ago
or, hear me out, organic machines have conscious experience because existence itself is divine. Humans don't have a special soul separate from the universe, they have a soul because they are the universe: materialism.
112.
by aoeusnth1 1y ago
That argument might make sense if Bing was as good, but Bing is worse. Wasting money building a bad product does not entitle you to market share.
113.
by aoeusnth1 1y ago
all that would make sense, but Bing is worse. Wasting money building a bad product does not entitle you to market share.
114.
by aoeusnth1 1y ago
LMArena is a joke, though
115.
by aoeusnth1 1y ago
This post is a very weak and incoherent criticism of a well formulated benchmark: task length bucket for which a model succeeds 50% of the time. Gary says: - This is just the task length that the models were able to solve in THIS dataset. W
116.
by aoeusnth1 1y ago
In principle, Math proofs are another relatively easy to verify problem. In the extreme case, you can express any math proof as a computer-verifiable formalism — no intelligence necessary. Step back one step, and you could have a relatively
117.
by aoeusnth1 1y ago
I have a hard time believing that he hadn't already made up his mind to make an open source model when he posted the poll in the first place
118.
by aoeusnth1 1y ago
Why not make the strong model compile a non-ai-driven test execution plan using selectors / events? Is Moondream that good?
119.
by aoeusnth1 1y ago
100 years optimistically!? That's an incredibly pessimistic timeline, maybe one of the most hardline "nothing ever happens" outlooks I've ever heard articulated.
120.
by aoeusnth1 2y ago
Their inference costs are the lowest in the business.

More ›