valleyer

3 ms·

1.
by valleyer 7d ago
"because duh"? OpenAI et al. have extensive infrastructure for running the model on a different machine from the one the harness is being run on, because... that's their main product. I would be absolutely shocked if the mo
2.
by valleyer 11d ago
Which people? This leak was caused completely by private businesses.
3.
by valleyer 12d ago
Looked at another way, the lack of grace period was indeed a problem back then, and that's why the Obama government proposed and finalized a rule to solve that problem. And now the Trump government proposes to reimpose the problem.
4.
by valleyer 12d ago
Another bit of evidence exposing the lie that Trumpists only object to illegal immigration.
5.
by valleyer 12d ago
Couldn't they use frontier open-weight models from Chinese labs? The current Chinese government is friendly to them.
6.
by valleyer 12d ago
Seems like it would be shut down by an injunction almost immediately.
7.
by valleyer 13d ago
You can't expect me to stop selling crack to kids on the streets. My job is to feed my family.
8.
by valleyer 13d ago
Not in vi. :)
9.
by valleyer 14d ago
At least in the US, police don't enforce terms of service of private corporations. That's a civil matter, enforced by filing a lawsuit.
10.
by valleyer 16d ago
Keep in mind there may be a significant Pepsi Challenge effect at play here.
11.
by valleyer 17d ago
Yes, I too remember when Shaq starred in this commercial. https://www.youtube.com/watch?v=Af0gtsjfy7E
12.
by valleyer 17d ago
Less space than a Nomad? Lame?
13.
by valleyer 18d ago
Funny. But... paper covers rock.
14.
by valleyer 20d ago
I'm not an expert, but... might this indicate the need for "sub-models", meant to be invoked by the general model to write good prose for it? Those sub-models would not be RLed on code (or any algorithmic-feedback tasks that
15.
by valleyer 22d ago
Prior art: the evil bit https://datatracker.ietf.org/doc/html/rfc3514
16.
by valleyer 22d ago
During training, certain tokens are more likely to lead to a lower loss function value, which is how you "win" the game of LLM output.
17.
by valleyer 22d ago
It could be added to context only on the inference server, with the model heavily instructed not to repeat its contents. Not foolproof, obviously.
18.
by valleyer 23d ago
"Domino effect" would probably be the more relevant named phenomenon there.
19.
by valleyer 26d ago
Interesting. I don't remember my crossover cables being marked that way; you had to read the dot-matrix printing along the side to know. Color would have been way more handy.
20.
by valleyer 27d ago
One of the strawiest strawmen I've seen in quite some time.
21.
by valleyer 29d ago
At least historically, it actually did run its own entire userspace, including daemons (all the way to launchd, IIRC), etc. All compiled for Intel, back in the day. Shared the kernel though. Many of the nominally shared system pieces betw
22.
by valleyer 29d ago
It is possible for both Trump and Biden to be wrong about something.
23.
by valleyer 29d ago
If punching random people on the street is so obviously bad, then why do people fight back when I do it?
24.
by valleyer 1mo ago
Anthropic would estimate (and provide justification of) how much lost income (and perhaps brand equity, opportunity cost, etc.) it suffered as a result of the government's actions, and the judge would consider it and decide.
25.
by valleyer 1mo ago
I agree. I just thought it was funny that someone was pointing out the legal unaccountability of an AI in this context.
26.
by valleyer 1mo ago
Good point. CEOs of companies are often held personally accountable for corporate malfeasance by "the current legal framework".
27.
by valleyer 1mo ago
Even if it's communication/announcements from the government ? I too am constantly frustrated that official communications from governments happen on Twitter or Facebook.
28.
by valleyer 1mo ago
Certainly. But in the current social environment, being seen as (mildly) careless is less bad than being seen as someone who lets an LLM write prose for them.
29.
by valleyer 1mo ago
"Reality has a well-known liberal bias."
30.
by valleyer 1mo ago
The U.S. constitution does not say that. Where are you getting that from? It does say that the president cannot pardon an impeachment, but that's not the same thing at all.

More ›