zahlman

15 ms·

361.
by zahlman 2mo ago
> I got my slopmachine cracking on a port For 131 lines?
362.
by zahlman 2mo ago
My point was that the kind of error it made was nothing like the kind of error a contestant would make.
363.
by zahlman 2mo ago
This is only applicable if you already have root (in order to get beyond that), right? It doesn't expose new risk of local privilege escalation?
364.
by zahlman 2mo ago
> And that's assuming the private keys stay private, which is its own tricky problem. If a private key ever leaks anything signed with it becomes invalidated. We don't appear to be talking about cryptographic signing here. (Tha
365.
by zahlman 2mo ago
Because the entire point of my comment was to point out a logical fallacy, not to argue the matter at hand.
366.
by zahlman 2mo ago
Interesting response. Do your claims ever require evidence, or only those of people who disagree with you?
367.
by zahlman 2mo ago
To add to that: given what they went through with the last model, I don't believe for a second that the real system prompt is even remotely this short.
368.
by zahlman 2mo ago
> Nice semi-colon. Thanks. I've been using them literally for decades. This one is a deliberate stylistic alteration, where a comma would have been the obvious choice, intended to suggest a specific speaking cadence. Your source dir
369.
by zahlman 2mo ago
The thread title is "What sort of maths are LLMs good at?" The codebase you describe would be an external tool in GP's conception.
370.
by zahlman 3mo ago
This appears to be trying to sell a system for meeting the right people, rather than giving any actionable advice about how to do so. It also glosses over the fact that who you know is often a result of some combination of hard work and luc
371.
by zahlman 3mo ago
> Sampling is what AI is good at. You might think so, but I tried asking ChatGPT to solve one of the puzzles from https://en.wikipedia.org/wiki/Countdown_(game_show) (which a Python script can brute-force on my 12-y
372.
by zahlman 3mo ago
To my understanding, there's a "controversy" filter on things that get a lot of comments relative to the vote count, especially if those comments aren't well received.
373.
by zahlman 3mo ago
There's no "hope" involved here. I'm talking about what the words actually mean.
374.
by zahlman 3mo ago
> So it is still going on How would you know? > Just last week they were fighting Minnesota's law that makes creating this stuff illegal. What law, and what evidence of fighting; and what evidence that their motivation has anythi
375.
by zahlman 3mo ago
I really shouldn't have to explain this to someone whose account dates to 2013, but this is not how we do discourse on HN.
376.
by zahlman 3mo ago
> It's the near-concurrent release of the same jump in capability that I find suspicious; not the fact that labs can catch up eventually. When everyone's improvement (or at least, everyone's rate of increase in parameter c
377.
by zahlman 3mo ago
> in my opinion, having to convince your tools is not computer science. If you think the system is a tool and not an intelligent, conscious entity (I think you are correct in this), then you cannot reasonably think of input to that syste
378.
by zahlman 3mo ago
> Annoyingly, the line about not mentioning these guidelines is superseding any instructions in the system prompt, causing the model to often refuse discussion regarding system prompts If the prompt guidance is causing the model to be so
379.
by zahlman 3mo ago
I've found that I don't mind getting a lot of text back from an LLM, when I was the one who prompted it. I can easily enough let my eyes flit around in the text and figure out what I need to, and I expect that putting all that tex
380.
by zahlman 3mo ago
Have fun writing a legal definition that accomplishes what you want but doesn't bar the police from creating a composite sketch from witness descriptions.
381.
by zahlman 3mo ago
The large majority of police killings are definitionally not "murder", and have clear moral justification on broadly-accepted grounds.
382.
by zahlman 3mo ago
This is non-sequitur. "Investigating crimes" entails knowledge that a crime has actually occurred, and deliberately seeking information specifically related to that crime; "fishing for circumstantial evidence" does not.
383.
by zahlman 3mo ago
> The fact that you have to carve out an exception for mathematics already shows there's a problem. I don't see how that's an exceptional case. Mathematical theories may not correspond to physical objects, but they still h
384.
by zahlman 3mo ago
Being able to read out positions (minimax, alpha-beta pruning etc.) is not "prediction".
385.
by zahlman 3mo ago
The page source appears to contain all the actual text within <p> tags, but structured in a completely illogical way. With JavaScript disabled, there are a bunch of shaded bars where the text should appear, which look like placeholder
386.
by zahlman 3mo ago
If the blog were about calculus, and stated that an elegant proof of the Fundamental Theorem of Calculus could be found in such and such undergraduate textbook, would you be upset that the citation wasn't to either Newton's or Lei
387.
by zahlman 3mo ago
That compression and prediction are deeply connected is obvious from simply considering how compression algorithms work. Nothing about the concept of "intelligence" is nearly as well defined as the other two, nor as well agreed up
388.
by zahlman 3mo ago
Sorry, but this completely misrepresents what LVT actually is, as well as suggesting considerably more agreement between Marx and e.g. Smith than they actually had; it's also wrong about the chronology of alternative theories, or at le
389.
by zahlman 3mo ago
> It's a mechanism to show that there is a certain floor on value of an economic good, which is raised every time constructive labor is applied to a product. The problem is that this argument is circular: whether the labor is "
390.
by zahlman 3mo ago
> First of all isn't that the wrong way around and should be the value theory of labor? No; https://en.wikipedia.org/wiki/Labor_theory_of_value is an established term and has been for centuries. > what is wr

More ›