daishi55

6 ms·

31.
by daishi55 1mo ago
How can you be so disconnected from reality? The valuations are based on the immense value of LLMs, which exists regardless of how consumers - who are very much not where the revenue is coming from - feel about it. I can see how much we are
32.
by daishi55 1mo ago
I don’t think you read the article, which documents many cases of him being wrong For example saying in 2024 that LLMs had peaked. That’s not early, that is already, definitively wrong.
33.
by daishi55 1mo ago
Feed the LLM output into a “deterministic” verifier, problem solved. That’s how LLMs verify their new mathematical proofs with lean.
34.
by daishi55 1mo ago
I think the idea that most tech layoffs since 2021 are due to Covid overhiring is lacking a grounding in reality and really any evidence at all.
35.
by daishi55 1mo ago
I suspect there are already plenty of test suites out there for something as universal as a networking stack.
36.
by daishi55 1mo ago
Ok so if it is indeed possible to verify the code after JIT generating it, I still don’t understand why you would assume that in the hypothetical scenario we aren’t doing that verification? You literally said it can’t be JIT, I’m just sayin
37.
by daishi55 1mo ago
I don’t know why, but it’s certainly possible to imagine JIT-generating code and then verifying it. Which is why I asked why we were assuming the verification wasn’t happening.
38.
by daishi55 1mo ago
Why can’t you JIT- generate the code and then run the verification?
39.
by daishi55 1mo ago
why are you assuming there would be no proper verification in this scenario?
40.
by daishi55 1mo ago
The robot overlords will put the last free man in a pod in 2080 and you guys will still be calling it Covid hiring correction huh
41.
by daishi55 1mo ago
Oh you sweet summer child
42.
by daishi55 1mo ago
Sure they have. Massive layoffs throughout the industry.
43.
by daishi55 1mo ago
> But are LLMs actually getting smarter, or just better at fooling us? They’re getting smarter. Next question.
44.
by daishi55 1mo ago
There was one proposal that focused on environmental issues with AI. This is indeed pretty disconnected from reality when you consider that Debian doesn’t have any policy requiring you to be vegan in order to contribute
45.
by daishi55 1mo ago
So is asking a human to write code, if you think about it…
46.
by daishi55 2mo ago
the word “genocide” does not appear in that article. Also, I was aware of it. Why did you think I wasn’t?
47.
by daishi55 2mo ago
Yes, those things are mutually exclusive.
48.
by daishi55 2mo ago
Sure. That’s possible. But that’s not what I was talking about.
49.
by daishi55 2mo ago
For OAI and Anthropic at least you can set a spend limit per response. Also tokens are well-defined.
50.
by daishi55 2mo ago
With how competitive the LLM field is, it would surprise me greatly if any of these players were doing anything other than trying to make the best possible product. I certainly do not believe they are intentionally training the models to us
51.
by daishi55 2mo ago
Absolutely! If Pakistan started committing a genocide and nobody talked about it but continued talking about Israel’s genocide, I would start to ask some questions. That is not the case though, no country on the planet is currently doing an
52.
by daishi55 2mo ago
Your original statement “LLMs use randomness in a fundamental way” is incorrect. LLMs have these knobs and randomness is not an inherent property of LLMs.
53.
by daishi55 2mo ago
It’s not “the Jews” it’s Israel, and I suspect the greater attention is due to the greater crimes being perpetrated by the Israeli government compared to those other states.
54.
by daishi55 2mo ago
I think the difference is probably related to the ends being pursued by each of those states at the moment. Only one is engaged in what the UN, HRW, and most legal experts call a genocide. So influence operations in support of an ongoing ge
55.
by daishi55 2mo ago
Claude has those knobs, they are just not exposed to the user. They could make Claude nearly completely deterministic if they wanted to (of course it would be a far inferior product then. But they could).
56.
by daishi55 2mo ago
Every cloudflare product I've ever used has been a legit delight to use. I use it for lots of stuff on my homelab. Everything works well and is simply pleasant. Maybe if you use it heavily for enterprise that changes, but it just hasn&
57.
by daishi55 2mo ago
I think that as soon as you start viewing certain groups/categories/etc as bad and blaming them for everything, as many engineers do to the non-engineering staff at tech companies, you have stopped thinking logically and are just
58.
by daishi55 2mo ago
We are using RISC-V for AI accelerators to great success https://ai.meta.com/blog/meta-mtia-scale-ai-chips-for-billio... RISC-V was a great choice due to being so customizable and extensible.
59.
by daishi55 2mo ago
Not directing this at you personally. But I've heard a lot of mediocre engineers whine and complain about PMs, MBAs, C-suite, etc. The best engineers I don't hear complaining much at all. Just a thought. And sorry but I don't
60.
by daishi55 2mo ago
What values and principles are supposed to be preventing cloudflare, the internet and technology company, from exploring hot new internet markets and technologies?

More ›