antonvs

13 ms·

271.
by antonvs 2mo ago
It’s the repetitiveness of style, the attempt to make everything seem as impactful as possible, the use of short sentences (too much Hemingway in the training data?), and obvious patterns like “it’s not this, it’s that” and several others.
272.
by antonvs 2mo ago
The terminology is misleading you. “Hallucinate” is misleading here. In the given example, a classification is being done very successfully - it’s just that it requires an extra step to map it to an arbitrary predefined list of classificati
273.
by antonvs 2mo ago
That’s one of the properties that distinguishes intelligence from traditional deterministic algorithms.
274.
by antonvs 2mo ago
Right, the point is it's a kind of elephant in the room - they even allude to it not being a common use case - but they say nothing about the rationale. That leads me to suspect it was a kind of "seemed like a good idea at the tim
275.
by antonvs 2mo ago
It seems surprising before you've learned and internalized it. It seems natural after that. I think this is a case of people not being very introspective about their beliefs about their work.
276.
by antonvs 2mo ago
You should take a look at "The Calculus of Constructions" (1988), which is one of the seminal papers that describes the basis for the Coq (now Rocq) theorem prover: https://www.academia.edu/47789825/The_calcul
277.
by antonvs 2mo ago
The standard Linux CLI tools are installed by a package manager, which has addressed the issues raised upthread. Installs via curl | bash don’t automatically benefit from that.
278.
by antonvs 2mo ago
> Not in any useful way, though. Theorem provers and proof assistants like Lean, Coq/Rocq, Agda, Idris, NuPRL and Epigram all fundamentally depend on Curry-Howard. And again, the point is not that there’s some magical isomorphism, b
279.
by antonvs 2mo ago
> That was hard to understand. It was LLM written or heavily LLM processed. This, for example, is 100% LLM: > Deciding boundaries up front was never about making the build easier. It was about making review possible, and it was the on
280.
by antonvs 2mo ago
Writing code and proving theories are flip sides of the same coin. See e.g. Phil Wadler’s “Proofs are Programs”: https://philarchive.org/rec/WADPAP The statement should probably really be “programs are proofs” since it
281.
by antonvs 2mo ago
> with hindsight come to seem beautiful and natural. …because they’ve been internalized. I wonder if the author has ever questioned where his notions of “beauty” and “natural” come from.
282.
by antonvs 2mo ago
Yeah it’s a pity that wasn’t addressed in the article. This is a little like “we shot ourselves in the foot and then performed surgery on our foot, and everything is resolved now.”
283.
by antonvs 2mo ago
We need a grassroots Tiktok campaign against Big Jellyfish
284.
by antonvs 2mo ago
That's the value of a single third-tier AI startup.
285.
by antonvs 2mo ago
That derailing is indicative of the mindset of the people who do it. All they see or understand is the immediate consequence of deaths of people, so they look for solutions to that problem without any consideration for broader systemic fact
286.
by antonvs 2mo ago
The reason I commented is that I faced that choice myself. I was arrested, went to court, and was given a suspended sentence conditional on reporting for service, and then left the country. Doing that had a major personal cost, but I wasn&#
287.
by antonvs 2mo ago
Beliefs need to be justified by evidence that supports a defensible theory. Otherwise, there’s no meaningful basis for choosing between beliefs. I’m not mocking you. I’m pointing out that you’re making an epistemic mistake, and speculating
288.
by antonvs 2mo ago
How are you accessing it? You may be using one of the limited models. I use Gemini Antigravity daily for software design and coding, mostly using the Gemini 3.5 Flash model. It’s very competent, and quite a bit cheaper than Claude.
289.
by antonvs 2mo ago
I tried asking a zip file to write a program for me but it did nothing. I’m starting to think that compression is not, in fact, prediction.
290.
by antonvs 2mo ago
He could have made the decision not to fight, he just didn’t want to. The idea that you “have to” fight is abdicating your moral responsibility. He chose to fight.
291.
by antonvs 2mo ago
LLMs reduce delivery time, and if you’re not taking advantage of that, you’ll discover first hand why it’s important. Would you also say that any message worth sending is worth writing by hand and mailing via the postal service, instead of
292.
by antonvs 2mo ago
> wtf am I doing Some part of you realized that Bronze Age superstitions aren’t worth that kind of effort.
293.
by antonvs 2mo ago
> No telemetry Must be tough not to be able to monitor your own models! (The odds that that tagline was AI-generated seem high.)
294.
by antonvs 2mo ago
That’s the position the EU is currently taking with telemarketers. What’s the issue?
295.
by antonvs 2mo ago
> If it didn't work, people wouldn't still be doing it. Yup. That’s why rape and murder is so popular.
296.
by antonvs 2mo ago
That doesn’t apply in this scenario. If CEOs start trying to emulate ethics in order to avoid being 86’d, we still get a good outcome. An issue arises when the metric isn’t a good proxy for the property being measured. But ethics would not
297.
by antonvs 2mo ago
> Once there is something, rather than the obviously more likely nothing at all Why is that "obviously more likely"? It's clear that you haven't really thought about this outside of the confines of your indoctrinated
298.
by antonvs 2mo ago
You're using "controversial" as a euphemism for "I'm basically MAGA now". "Rolling eyes" doesn't come close to expressing what disdain I have for your position. Educate yourself even minimally on
299.
by antonvs 2mo ago
What happens if we do the same for CEOs?
300.
by antonvs 2mo ago
When Zuck’s not on board, that ship is effectively his. I bet he stands at the bow yelling “I’m the king of the world!”

More ›