baq

15 ms·

301.
by baq 4mo ago
the possible futures after the thing is built are uncountable, but hoping the thing won't get built at this point is naive. in general I agree people should be reading a lot more sci-fi nowadays than they used to.
302.
by baq 4mo ago
basically imagine the Manhattan project, but instead of blowing up the desert they're building the biggest datacenter you've ever seen.
303.
by baq 4mo ago
that's... an optimistic take I think
304.
by baq 4mo ago
who is going to continue to publish these open models and why would they keep doing it?
305.
by baq 4mo ago
the whole thing playing out as expected. if you think about it, the only question is the timeline. the next model with a gap to mythos as mythos is to opus will be controlled technology from the get-go. the one after it may be top secret.
306.
by baq 4mo ago
this is exactly the play is my point
307.
by baq 4mo ago
> Regulatory capture is the OpenAI and Anthropic end goal, for certain. it has to be, because the other way around - the government taking over parts or the whole thing - is inevitable if the trend holds.
308.
by baq 4mo ago
what Dario wants is to retain any influence whatsover on how the research progresses before the inevitable nationalization of the frontier. he gets to keep the N-2 tech and maybe influence the N-1 tech, but the only influence on the frontie
309.
by baq 4mo ago
What do you mean by ‘idle’? If you mean your database is seeing extended periods of no updates to a table, you still want to vacuum, maybe even vacuum full if you know when traffic stopped and for how long to get the best possible read perf
310.
by baq 4mo ago
I’ve been piloting frontier LLMs for as long as anyone outside of the labs and I just disagree. It is a tier above for some tasks (especially in my usage) and not a downgrade on anything I tried it on. This is enough for me to rank it highe
311.
by baq 4mo ago
It is on a level above everything else for now, that’s enough to determine it’s quite literally in its own class. Anecdotally it is a good model, sir.
312.
by baq 4mo ago
Yes, pretty much, except getting to working fusion is not guaranteed - but the climate disaster is, so we have to make a bet. Solar and batteries are the obvious fallback plan. Nuclear should also be, but it’s a hard sell.
313.
by baq 4mo ago
DeepSeek v4 and GLM 5.2 are not Mythos-class, the capability uplift as measured is continuous but consequences are step functions.
314.
by baq 4mo ago
No idea, ask fusion scientists, they might have answers on what they need built a lot of and quickly and whatever it is it’ll take energy to do it. I wouldn’t mind shortening e.g. the Iran war by a couple of days and redirecting the funding
315.
by baq 4mo ago
You should run vacuum as often as possible in Postgres if you’re doing anything other than INSERTs, this is a design tradeoff in Postgres itself. It’s the reason autovacuum exists and why tuning it is so important for performance; nothing w
316.
by baq 4mo ago
> should have an open source Chinese model within 6-12 months that's already end-of-world level that's the exact thing I'm talking about. I don't see why is half the people around here so sure that China will continue
317.
by baq 4mo ago
stop listening to politicians, start listening to scientists maybe, 15 minutes of thinking isn't good enough for anything
318.
by baq 4mo ago
the challenge in carbon tax isn't the people who vote it out, it's the people who never vote it in in the country next door (or on the other side of the world, it hardly matters)
319.
by baq 4mo ago
We can't slow down burning stuff for energy, this is politically untenable. ...so the answer is to accelerate the burning, but not for the sake of burning more, but to focus on getting to true clean energy sources which will allow us t
320.
by baq 4mo ago
‘Why wouldn’t anyone give away frontier AI?’ sounds like ‘why wouldn’t anyone give away uranium enrichment?’ i.e. I can’t comprehend the state of mind and the world model of anyone asking a question like that, which is apparently quite a fe
321.
by baq 4mo ago
See also https://andrewmccalip.com/space-datacenters with some models and numbers you can play with.
322.
by baq 4mo ago
This valuation is derived from the fact the three tokens aren’t free
323.
by baq 4mo ago
Do you complain to the bank that credit cards have expiration dates or to the government that passports also do?
324.
by baq 4mo ago
Immediate revenue impact is basically 0 - nobody cancels their Claude sub because Fable isn’t why they got it in the first place (by nobody I mean like 1% of total users and they’re likely net neutral tokenmaxxers for revenue). Signal to Op
325.
by baq 4mo ago
Burning weights onto a chip in an efficient way and exposing that via USB would be acceptable for a good enough model tbh
326.
by baq 4mo ago
For decades we were encrypting our communications with rsa, surely nothing is wrong with it?
327.
by baq 4mo ago
Organized violence is self regulating if it wants to sustain itself.
328.
by baq 4mo ago
The fallacy of the libertarian argument is the assumption that it’s possible to have a small distributed government trivially doesn’t hold in presence of bigger adversarial monolithic governments elsewhere.
329.
by baq 4mo ago
Except the Israeli versions obviously. These are at least on par with domestic.
330.
by baq 4mo ago
US has some questionable allies themselves who happily and remorselessly stole top secret information including nuclear secrets.

More ›