zahlman

6 ms·

1.
by zahlman 3d ago
> Wouldn't the simplest solution for stopping the proliferation of wanton felony generators just be holding operators liable for actions that their agents take? What happens when we end up with effectively a botnet of wanton felony
2.
by zahlman 3d ago
I find it implausible that they simultaneously are so worried about these recursively self-improving systems that they don't properly understand, but can freely "invisibly fuck with the weights" on that level.
3.
by zahlman 3d ago
Pangram's algorithms are not open-source, so I wouldn't expect this to work very well.
4.
by zahlman 3d ago
"Holding people accountable" is not going to help matters in the event of what amounts to simultaneous terror attacks on basically every system. Yes, a single LLM without tools isn't doing anything but writing to standard out
5.
by zahlman 3d ago
"Going to" can't be said, but there are very serious risks. The problem is the difficulty of scaling back agency for these systems in a world where the Internet is so ubiquitous. Not only do we expect basically everyone to be
6.
by zahlman 3d ago
Just because a proof is valid doesn't mean it proves the thing you want it to prove.
7.
by zahlman 3d ago
I'm confused. You started out with "That's like saying", but you appear to agree with me completely.
8.
by zahlman 3d ago
> but it's pretty likely that "optimized for humans" and "optimized for agents" are not identical. Why? > we really don't need people to be in the code Code exists for humans, if only for safety reasons.
9.
by zahlman 3d ago
> That there was zero barrier to building anything anymore. ... Surely this would mean we would get new ideas faster if that was the case? No, of course not. The point is that the tools comprehend the idea and implement it. Not that they
10.
by zahlman 3d ago
> It suggest that some sort of borrowing takes place It does. A value is passed by reference, borrowing it from the owner. > and that it's just an optional check I don't see how the word "check" implies that it&#x2
11.
by zahlman 4d ago
Sure, but it would be actively bad to make the code more complex simply because we have machinery that helps us deal with the complexity. A big part of how people assess the models' coding capability is whether they create needless, in
12.
by zahlman 4d ago
> Consumers literally don't know the difference between cellular/LTE carrier "5G" marketing and the "5 GHz" band on their home router now. Why would they know the difference? They don't have any technic
13.
by zahlman 4d ago
This is the first I've heard of any plan for 6G….
14.
by zahlman 4d ago
The comment concludes with "Conclusion: no, bad advice, except for very simplistic scenarios.", so I don't think I can agree with you on this. My point was that these facts don't (in my analysis) support the conclusion,
15.
by zahlman 4d ago
> If you have a swap file created by the installer then you do not get to specify Why should we care about what the installer does when it's trivial to change later? You said this is "bad advice, except for very simplistic scen
16.
by zahlman 5d ago
> If you have a small root partition, the default config Of what distro? If you mean that the example code shows the swapfile directly under / , well of course you don't have to do it that way. > With a partition, it's
17.
by zahlman 5d ago
> between the two calls, an attacker could open the file, and then keep it open until next reboot. I feel like if there's a malicious user (or program) on your system with the necessary access for this, you have much bigger problems
18.
by zahlman 5d ago
"Hug", not bug. Referring to a site (presumably under OP's control) crashing due to increased traffic from HN.
19.
by zahlman 5d ago
I think GP is about PII rather than "mutable user data".
20.
by zahlman 5d ago
> Stereoscope — Toronto 2008, City Hall: two towers, one matrix (96×32 pixels, 16 greyscales). Oh man, I think I remember that.
21.
by zahlman 5d ago
> fully public after like 20 years unless releasing it would put a specific person or named group in danger As the recent AI safety discourse should make clear, the set of information that potentially endangers people could be much broad
22.
by zahlman 5d ago
> Ultimately we think a fatal flaw was found in Mochizuki's proof, so it didn't lead anywhere in particular. But in our hypothetical "AI lean-verified proof of RH" situation, it would presumably generate substantially
23.
by zahlman 5d ago
I mean, I usually check the weather by putting "weather" into a search engine. And this will show me little SVG icons that take up a tiny fraction of my low-resolution screen and probably involve like less than a kilobyte of path
24.
by zahlman 5d ago
It feels like I keep ending up leaving a comment on the unpopular version of a submission that ends up being reposted and getting traction later. https://news.ycombinator.com/item?id=49553744
25.
by zahlman 5d ago
> I know you asked an AI later and it told you about hash collisions but I'm wondering where this first comment came from. Was it also AI? That was two different people. The first comment probably comes from finding the 1<<61-
26.
by zahlman 5d ago
> It emits groups of five characters. Five is the sweet spot: 84⁵ = 4,182,119,424, only 2.6% short of 2³². > That leaves enough room for a group to hold 32 bits about 95% of the time on uniformly random input, and 31 bits otherwise. &
27.
by zahlman 5d ago
It seems hard to believe they could expect to get away with this, given model-to-model differences in writing style.
28.
by zahlman 5d ago
> you still need internet connection Well, yes? Did you expect a local model?
29.
by zahlman 5d ago
Why would you need that kind of resolution for a weather app? What are you even going to depict with all those pixels?
30.
by zahlman 5d ago
> It sounds like a pipe dream, but this is how physical copies have worked for decades. My license moves with the game disc or cartridge. Why don’t we have a digital equivalent to this? FWIW, everything I ever bought digitally in the Nin

More ›