chermi

How did they pick winners and losers early days of Tesla? It demand side subsidy. No one else could meet it. It worked.

Wow having it as a PCI card would be awesome. I don't know the feasiblity but it would be really cool. Ordering different models, back to physical media basically...

weird, I'm pretty sure there was some american that made electric cars a thing. I'm not a fan of state sponsorship or bailouts. but if you're trying to equivocate, it's just not really true. also anti-tarriff. I guess I don't know who you're arguing with.

edit-- and re industrial policy, I'm ok with demand side stuff like government contracts in the early chip days. Less so but kind of ok with some supply stuff like EV credit, but of course in that case would've preferred the politically impossible carbon tax.

I never said it wasn't competitive, I said I'm not optimistic. There's lag effects to everything and I think we're just seeing the beginning of it coming due.

My gut is that, at least for now, there's a timescale mis-match problem. Hardware still takes too long, then you have to deploy it. I don't know much about "burning asics", but if the whole process of spinning up programmable GPU data centers is months, I imagine the whole ASIC cycle has some catching up to do. To be clear, by timescale mismatch I mean model quality improvement timescale vs. deployment timescale. But maybe we're finally getting to the point where behind-the-frontier-but-cheaper are in sufficient demand and ASICs make it cheap enough to close that gap.

There's a difference between state sponsored and private dumping.

Dario's PR strategy is one of the most confusing things I've ever witnessed.

It's basically American VCs vs the China the state. I'm not optimistic for the US at this point, given how much China cares about it and how much talent they have. And how much they're putting into hardware and the whole ecosystem. Meanwhile we have pro basketball players with no understanding of reality being celebrities for decrying data centers because...land?

I mean isn't the explanation simply that llama was never good enough, even when it was released? I hear (no data) lots of people using gemma4, at least a month or two ago.

I think I am, but I wouldn't really call it offloading thinking. I would call it trying to offload thinking

Did you mean landauer?

Late to this post, but my impression was that later models would be more efficient per task? Wouldn't they save compute released fable 5, maybe capping the effort, if it is actually a better model?

Can you please point me to the proof of the first claim?

Lol I feel like no one has any attention span here. Tech shit is expensive in the beginning when it's new. It gets cheaper with time. This is a tech forum, don't we know this? Of course people overreact in both directions on both sides of the issue. It's a very fast technology, wait for things to settle before making grand declarations.

I mean, didn't we give the government and the public long enough to prove they could provide abundant, cheap nuclear? They were so closed in the 60/70s and have since failed miserably and everyone has suffered for it. Cheap, abundant energy is good for humanity. If a private company accelerates it, I'm here for it.

Yes, I'm also for solar, and wind, and geothermal, and nat gas, and way out there fusion. It's hard to exaggerate how much cheap, abundant, reliable energy helps civilization.

That's the permit/approval for the pilot/test, right? There are about a million approvals they need to get through. Are they using the DoE fast tracking method?

Yes. And to anyone paying attention, this has been current since about 2010.

Because the laws are different? Are you really confused?

Most mathematicians don't take pride in their results having no applications. That's just not true. Maybe some quirky pure logicians or something. But otherwise 90%+* of mathematicians I know would be at least satisfied if not thrilled for their work to be used by others.

*Completely made up statistic.

This is why he needs a down vote button

Wouldn't that just accelerate collapse? How much do you trust the outputs of the llm to provide trustworthy and valuable new information? I mean I understand distillation works. But that's much more structured and thoughtful than my sessions at least.

Ummm, why not both?

I am very skeptical that musk is 10-20% interest. I would guess closer to 5.

I've accidentally clicked ai mode probably 3+ times a week recently, so that's some real good metrics ;)

If only there was a way to think beyond direct substitution.

More predictive power is always a good goal, full stop. This is orthogonal to whether the model producing prediction helps with "understanding" directly. Predictability encodes understanding in a strict information theoretic sense, regardless of our ability as humans to access that understanding.

Per frontier token. You're not calculating the cost of a fixed quality asset here. Old hw running non-frontier models will be very valuable. In fact, we have two direct examples: older server gpus actually appreciating and the very obvious fact that not everyone always use MAX FULL EFFORT BEST MODEL no matter what.

What? Go volunteer at a botanical garden or something.

Yeah. It's called brain drain. Talent has options. It weighs pros and cons. When the relative attraction of a country and thus institutions within it drops, they choose to go there less.

To be clear, I would still choose to do my PhD in the US. But this is a marginal effect, people weigh many factors. If you think, for example, you're going to be constantly worried about visa issues, you may just choose Europe or China over the US.

Edit- sorry NZ and australia, forgot about you

It's really kind of gross. Psychologists should know best about what kind of damage the social media shit does.