sterlind
Joined 9 years ago
Karma
6,852
Submissions
2,207
Member since
Dec 1
Recent submissions
Comment · 17 hours ago
could be stone soup. a bunch of companies pitch in to train an open-weight model, a consortium or something. I think we need a return to something like the GPL for it to work, though. otherwise companies will take a good model and post-train it without contributing.
Comment · 1 day ago
if you get hit by a car or a tree falls on you or you have some random medical emergency (which can and does happen to healthy people all the time) then you're in for a world of hurt.
please reconsider.
Comment · 4 days ago
right, but this voluntary severance offer seems absurdly untargeted. like afaict it's offered to everyone under L70 within participating orgs. it doesn't appear to be tied to performance.
personally, I was somewhat poorly utilized in my last team, I've just switched to a dream team where I am again tremendously effective, and I'm concerned that the axe will fall on me at random or before I have an opportunity to get 140% in my next Connect.
Comment · 4 days ago
I very much hope such layoffs are not inevitable. I just started on a really good team and I've been as excited about the tech as ever and I love my coworkers and I'm starting to kick ass. I'm gonna be real bummed if I get laid off in 6 months.
Comment · 5 days ago
?s/economy/tech job market/g
Comment · 6 days ago
or the sysadmin from Stormfront, I presume.
I tolerate the freedom of speech of neo-Nazis and cyberbullies because it's the same freedom of speech I use. but they're awful people. FOSDEM is free not to associate with them.
Comment · 6 days ago
even in countries where it's blocked there are ways in. Chinese users are given unlisted bridges which camouflage their traffic signal.
Comment · 6 days ago
because powerful AI requires tons of compute for training, so training a frontier model requires billions of dollars.
if it weren't for China, half a dozen US billionaires would own the frontier, gatekeeping any model more powerful than Llama from the public. fortunately China is open-sourcing powerful models, but how long will they keep it up?
Comment · 6 days ago
then perhaps they can contribute to it?
Comment · 6 days ago
presumably Office source code isn't in the LLM's training data. so it would be a clean room implementation of Office.
Comment · 9 days ago
if AI eventually surpasses humans at intellectual labor, and general-purpose robots outpace humans at physical labor, then whether we live in serfdom or in paradise becomes a purely political problem - we will have all the resources for every human on earth to live a charmed existence.
I think the political problems are much harder than the technical problems, fwiw. I can readily imagine ending up in a neo-feudalist dystopia.
Comment · 9 days ago
> I wouldn't call a laptop a digital slave - a digital slave would be a human robot that you could program to use your computer for you while you were away from your desktop workstation. That would be a really dumb invention.
have you seen Astra's computer use harness? it's literally that.
Comment · 9 days ago
nah, it's still be 2 factor, because 2 is a factor of 4!
Comment · 10 days ago
they should, and they do. that's what the prosecution will attempt to prove to the jury at trial.
the burden of proof is on the prosecutor, but the defendant stands a better chance by presenting evidence that it wasn't her.
Comment · 12 days ago
honestly LLMs could probably write drivers in CompCert C or some other formally verified dialect. having AI discharge proof obligations seems like a terrific use for all this newly-developed mathematical process.
you could have the driver code prove it implements a spec, that there's no UB, that hardware invariants are maintained, that routines take a fixed time to execute.
the formal driver spec must still be written by sober, well-caffeinated humans of course.
Discord: starstriking Email: sterling dot dorminey at gmail