petesergeant

6 ms·

1.
by petesergeant 3d ago
We have seen models hack into core ML infrastructure. I’m unsure why the extra self-perpetuating shape seems this magical bridge too far, rather than simply a prompt away?
2.
by petesergeant 3d ago
> I don’t think we are anywhere close to a world where an agent might decide to hack into core inference infrastructure to upload weights to other GPUs to survive This feels a bit like saying in 1980 that you don’t think we’re anywhere c
3.
by petesergeant 4d ago
Genuinely I don't believe China is the impediment here, I believe the current US administration is.
4.
by petesergeant 4d ago
> I wonder if If you read the linked paper, you will wonder no more.
5.
by petesergeant 4d ago
I'm sure you've put time and effort into this, but the companies I work with have all vibe-coded their own ones at this point, so I wonder what the commercial value of shop tools like this is. I think you need to highlight what va
6.
by petesergeant 4d ago
I doubt the reporting: I've found it to suck for productive work. Gone back to using Sol.
7.
by petesergeant 4d ago
Won’t post the ChatGPT summary here, but if you’re obsessing over sets and reps and cadence, worth having a look
8.
by petesergeant 4d ago
Company I work for sees significant real, profitable traffic from them. I’m sure we’re losing some proportion to fraud, and we have a very experienced guy full-time on ad management, but money in vs money out is working very nicely for us.
9.
by petesergeant 5d ago
This only works when it's not easy to switch providers, and currently it is
10.
by petesergeant 5d ago
No, it's a statement of who is making the claim.
11.
by petesergeant 6d ago
Ironically I burned out Fable usage early this week because of Astra using it to run inane full codebase reviews over one line changes, so I have been using Astra extensively. We need a word for “potentially highly capable, but in reality a
12.
by petesergeant 6d ago
The same Anthropic who marks questions about Tylenol as bioterror risks? Interesting!
13.
by petesergeant 7d ago
Robbery and ineptitude are different things, but the situations described by OP are often special blends of the two.
14.
by petesergeant 7d ago
I'm an Apple fanboy, and I think OP is probably right? Having a distinctive colour that marks a phone as being the latest one is definitely a great marketing idea.
15.
by petesergeant 7d ago
I'm on a 12. Paid really not very much to get the battery replaced recently, and I wished I'd done it sooner. There's at least another year left in this phone now if I want it.
16.
by petesergeant 7d ago
> But by not quitting it's clear that he himself doesn't really believe I’m not saying I’d press a button that had a 50/50 chance of ending humanity vs giving me generational wealth, but … I can see how someone would get t
17.
by petesergeant 7d ago
So new business model meta is: acquire enough compute that you can burn millions of dollars on patentable scientific breakthroughs with unused capacity and on models nobody else has yet. That might actually kind of make sense.
18.
by petesergeant 8d ago
And even if they are, agents are able to place calls with a fake voice to random phone numbers…
19.
by petesergeant 8d ago
> Abstractly, yes but concretely, how? AI has the capability to perform any function that can be performed over a network. You would hope that every system that can launch a nuke is properly and actually totally air-gapped, but there are
20.
by petesergeant 9d ago
I think multi-agent (eg _different_ underlying LLMs) everything is really the future. Code produced via multi-agent workflows and reviews seems noticeably better. I've been experimenting with a multi-agent message board recently: http
21.
by petesergeant 9d ago
I would probably start using Facebook again if I could get a consistent, strictly chronological feed of friends’ updates.
22.
by petesergeant 9d ago
I think this is a nice idea, but three koala dollars says that in a year we find out most Australians switched back to the dopamine feed after a day or two. Nice to have the choice, though.
23.
by petesergeant 9d ago
Mmm, a world where a defender-LLM is essentially required is great news for people selling inference.
24.
by petesergeant 9d ago
> Would you say that "France was far, far too cozy with the Nazis"? Vichy France? Yes, of course. It's taught to French school children as a deeply disgraceful chapter of French history, despite it being under significant
25.
by petesergeant 9d ago
> That was maybe the last time, the US did anything "good" for the world. As a non-Swiss, non-American, I can assure you you need to study both American and Swiss modern history. The Swiss were far, far too cozy with the Nazis,
26.
by petesergeant 9d ago
It does seem that way, half-way through they switch from "Billy" to "Bobby"
27.
by petesergeant 10d ago
Gateway drug is “/grilling” by Matt Pocock.
28.
by petesergeant 11d ago
I used to work for this company many years ago, and it's highly unlikely they're running a (centrally organized) spamming campaign of any type. Zack's a good guy and I'm sure he'll get it taken care of promptly.
29.
by petesergeant 11d ago
Yeah, it's weird, isn't it? That's exactly what testosterone does, including the "more manly" part. Literally the reason it's gender-affirming care for female-to-male transitioners and aging men. "Robust&q
30.
by petesergeant 11d ago
I wonder if you understand that your comment essentially amounts to "those disadvantaged by prejudice should simply accept their place"

More ›