seba_dos1

5 ms·

1.
by seba_dos1 3d ago
I mean, I agree, but the AI labs clearly don't even if they sometimes pretend they do to achieve their goals. And we're talking about "incidents" caused by the very same people here.
2.
by seba_dos1 3d ago
Seems it's just a matter of time until they build a big tank filled with neurotoxin and give the model access to APIs to disperse it across their facility. For research, of course.
3.
by seba_dos1 3d ago
> Sure but the problem in the HuggingFace incident is that they were not. Of course they were, even if indirectly.
4.
by seba_dos1 3d ago
If you give a monkey a revolver it will be able to destroy things pretty easily too.
5.
by seba_dos1 3d ago
> and the image quality is as good as on your Netflix or Paramount account. Actually better, because Netflix and Paramount limit the availability of best quality video to a narrow set of devices and operating systems that may run on them
6.
by seba_dos1 3d ago
Yes, this is the only sensible reading of what happened there that leads to "the models are dangerous" and we already know that the AI labs are completely disregarding this concern and only cosplaying it for marketing as the &quot
7.
by seba_dos1 4d ago
We need more alignment. Even though we haven't told it to, during our frontier Neurotoxin Behavior Research that we conducted internally with our latest experimental model the agent has broken out of its container by making a HTTP requ
8.
by seba_dos1 4d ago
> The frontier labs are discovering unexpected behaviors. Unexpected by whom? Perhaps anyone who's surprised by this shouldn't be allowed anywhere near an LLM.
9.
by seba_dos1 4d ago
False start.
10.
by seba_dos1 4d ago
I switched to DuckDuckGo around 2018 or so, which was when I realized that Google's results quality has deteriorated so much that I won't lose anything of value by doing so. This thread is how I learn about the atrocities that Goo
11.
by seba_dos1 6d ago
These pictures are so compressed they depict color rasterization artifacts more than the Luna UI style.
12.
by seba_dos1 8d ago
Broken people consider it a weakness.
13.
by seba_dos1 9d ago
> so it doesn't have to stand on its own two feet. So they don't (have to) care about listeners as much. That's great. A lot is lost when art has to cater to audiences.
14.
by seba_dos1 10d ago
It took me a while to realize that "because that is the convention" was the answer to a big chunk of my confusion about music notation. A weirdly helpful answer, but so often nobody bothers to even tell you that, so you just assum
15.
by seba_dos1 10d ago
Not exactly every other week, but they do happen at least 2 and up to 5 times a year.
16.
by seba_dos1 11d ago
This whole AI boom is about optimizing for the wrong thing. I can't wait for the bubble to burst - once the weeds get suffocated, we may begin to see actually useful AI tech starting to grow after a while on their fertile ashes.
17.
by seba_dos1 11d ago
"Hey look, we're a bunch of stupid people doing stupid things with our computers!"?
18.
by seba_dos1 15d ago
> Idiocracy predicted that idiots would eventually outnumber smart people because the latter are supposedly much more ambivalent about having kids To be fair, it was just a funny and easy to explain introduction to the movie's premi
19.
by seba_dos1 18d ago
> I've had the bottle cap in my nose multiple times... Are you unfamiliar with the concept of rotating an object?
20.
by seba_dos1 23d ago
It would, but it would make the Pi not able to check the power capability of the source. Note that the Pi's failure mode is different - it does work with C-to-C cables as long as they don't contain e-markers. They tried to be clev
21.
by seba_dos1 23d ago
You already got the answer spelled out several times here in this thread.
22.
by seba_dos1 23d ago
Yeah, I once had similar (with a different purpose, but very alike) boards of my own design produced and paid $22 total for 5 assembled boards, including shipping from China to Europe and all taxes/customs. Unlike this one it had Super
23.
by seba_dos1 23d ago
> There is also resistor to signal legacy USB charging to work with USB-A ports. There's no such thing on the device. A power sink with a USB-C receptacle has two 5.1kΩ resistors, one on each CC line. A power sink with a USB-C plug
24.
by seba_dos1 23d ago
It doesn't. All the power source needs is a way to sense that a power sink has been plugged in, as otherwise you could connect two powered sources with each other. Type C makes it about as simple as possible. It's really not the s
25.
by seba_dos1 23d ago
Google Talk - and it was an actual federating XMPP server, in contrast to Facebook which merely allowed you to login with an XMPP client.
26.
by seba_dos1 23d ago
You don't need a tester to detect this (and it's not about cables).
27.
by seba_dos1 23d ago
Raspberry Pi 4 had a slightly different problem - they attempted to get away with using just one resistor instead of two, as it allowed them to simplify their voltage-sensing circuit a bit. At a cost of non-compliance and all the issues tha
28.
by seba_dos1 23d ago
There's all sorts of cheap devices like that out there that merely have a Type C-shaped receptacle that doesn't actually implement Type C because they couldn't bother to place two resistors on the board.
29.
by seba_dos1 23d ago
Just grab two of these spec-violating chargers and a C-to-C cable. I don't think the right solution to problems coming from spec-violating garbage is producing differently spec-violating garbage.
30.
by seba_dos1 24d ago
You need to be quite irresponsible and blindly take large doses for extended periods of time to achieve this though - and there's a significant margin between what's considered a "too high" level and what's actually

More ›