eigenspace

Nearly 70% of eligible Americans didn't think that Trump was a enough of a problem to vote against him in 2024.

This betrays a complete lack of civic seriousness, and a childlike belief of invulnerability to consequences.

If you want to run reasonably big, local AI models, what are your alternatives?

That may not be many people, but there certainly will be some people who want to do that, and are willing to pay big bucks to do so.

> Every basis point of artificial yield suppression is a subsidy to procrastination.

Trump's entire modus operandi for his entire career has been to heavily borrow against the future to chase short term benefits, and hope that the consequences can be tossed on someone else's shoulders.

I should have been more clear. While distillation is part of how we got lucky here, what I really think is that it's just surprising and lucky that such a heavily data-driven approach ended up being so powerful here.

Transformers are like just a step or two removed from being fancy convolutional neural networks. I guess I'm just surprised that it didn't turn out to require more 'special sauce' with extremely elaborate internal architectures, and less of a big-data approach.

Because the data is so central in building these LLMs, rather than some special insights or ideas in the model architecture, or very special hardware requirements, the field is much more open than I would have guessed some years ago. And it's the fact that the data is so central that makes distillation possible in the first place.

Even if the transformer paper wasnt published, the info would have diffused out eventually.

Its not that wild of an idea. It's not like e.g. chipmaking where even just knowing how things are done doesnt mean you can copy it.

Im sure that's well on its way.

Jeeze, what are you doing that uses so much electricity?

I live in Germany where people won't stop whining about electricity prices, and I pay 75€/mo.

Yes, but 10 or 15 years ago, I would have thought that there'd be more to it than just a slight modification on the ideas behind a CNN to get this level of AI.

There were somewhat good reasons to think it needed more than just this data-driven ML approach.

I wouldn't be so sure about that. The popping of a bubble is usually just as irrational as its rise.

If investors start fleeing from senseless businesses in the AI sector, that does not mean that sensible businesses will be spared. These things follow herd mentality, and the primary drivers of the herd are greed and fear, not fundamentals or business logic.

I'd take a market crash over a monopoly in the hands of a ghoul like Altman.

The economy he and his ilk want to build is infinitely worse.

The fact that AI models can be so easily distilled and replicated is such a stroke of luck.

10 or 15 years ago if one had asked me to envision a future where a private company invents artificial intelligence, I'd have thought for sure they'd have a massive moat, be very difficult to catch, and it would create an almost instant monopoly.

Rather, it seems that selling intelligence might end up as a race to the bottom.

Who woulda thought that just having access to enough textual inputs and outputs and a vaugely similar transformer architecture would be enough to copy-cat rather useful intelligence.

It's not really about learning to rank yourself against other people's bodies more or less accurately. The point is to stop obsessing so much over how your body compares to other peoples' bodies in the first place.

My experience after visiting nude saunas and beaches in Germany is that you're in an environment where no matter what someone's body looks like, the social responsibility is to be casually respectful, and not make a big deal out of someone's body regardless of whether it's normal, or exceptional in a positive or negative sense.

This helps one put into context that it doesn't actually matter so much what your body looks like. You can be beautiful or hideous, and still have a right to relax and feel comfortable and go about your day.

I think that clothing has a way of amplifying our curiosity and reinforcing taboos. When you strip away the clothing you learn to relax about this whole business of fixating on how your body might compare.

Oh wow, I missed that they had spoken out on this. Yes, if that's real, it is truly disappointing.

We still haven't seen what the EU's reaction is to Apple's latest app store stunt.

This would not be the first time that Apple unilaterally claimed something was DMA compliant, and then the EU had to come out and say "no, that's clearly not"

While I agree I wish this would happen faster, it certainly is happening, and it's happening pretty quickly. Saying it "couldn't be further from the truth" is just silly hyperbole.

Rheinmetall's current output is 1.1 million per year, aiming to hit 1.5 by 2027 and continue to scale up from there. That is just one company in one country in the EU, though they are the largest one.

The EU also simply has way more capacity to scale up shell production than Russia does, and the EU is building up stockpiles while Russia uses up everything they make.

You are not very informed on the EU's artillery shell production capabilities.

If anything, the major gap is the EU's lack of cheap drone manufacturing, but even that is starting to make progress with help and integration from Ukraine.

They paused some investigations while the negotiations happen, but most of those are back on track now.

Yes, the EU does rely on buying US gas as an important part of its energy mix. But that is true independent of whatever deal the EU struck with Trump. The EU simply does not have control of who its gas importers buy from (other than sanctions)

But this is also why the EU is aggressively pursuing and building up alternatives, including investments in Canada's ability to export LNG, but also buildouts of solar, wind, and batteries.

Solar and batteries come with their own dependencies on China, but at least those are a different sort of dependency than fossil fuels. If China stops exporting solar / batteries, the ones we have will still keep functioning for decades while alternatives are built up.

The EU would strongly prefer to have US military support, but does not fundamentally need it. In the miltiary domain, but also lots of other domains, the EU is using these fake promises they made to Trump as a mechanism to buy time while they quietly work on removing US dependencies, and reducing US leverage.

That's what I was talking about with the false promises.

The EU has no mechanism to make those investments in the USA, and no mechanism to force any member state or company to make investments.

They literally just made up a number and said "oh yes, we'll make a bunch of investments" with zero intention of doing anything to make it happen.

The EU didn't really fold, they blew him off. It was just a bunch of "yeah sure Donny, here's a bunch of meaningless promises we have no ability or mechanism to actually enforce". For example they promised to buy a bunch of gas, but the EU has no mechanism to make companies or even member states buy gas.

The only real concession the EU made was just allowing Trump to tariff them at a certain rate, and not counter-tariffing back (with various escape clauses).

While the EU response certainly wasn't a feel-good display of chest thumping, I think it was somewhat effective. He was able to strut around and feel big, and they were able to move on.

The only reason the EU would really want to apply counter tariffs against the USA would be to either make a point, or because they are actually afraid that exports from the USA could outcompete EU products unless tariffed, but this just isn't really the case right now.

___________________

The only form of retailiation I think the EU should actually be pursuing is the equivalent of tariffs on US software. This is the area that the USA is a highly competitive 'exporter', and putting up barriers agaisnt these software services would be a good way to hurt the USA, and protect / stimulate domestic alternatives.

Oh I hadn't noticed, but youre right!

They briefly trialed a feature that told you how many reviews for a restaurant were deleted in the past year, and that was quite nice.

I hope it comes back.

This is part of why VW now contracts Rivian for their user-facing software after their own in-house software daughter company failed.

Bad timing though, in an age where talking about software sovereignty is all the rage.

I get more value from my smartphone than I used to, and I upgrade it less often.

The photos I take with it are genuinely precious to me, and I'm more than happy to shell out 800-1000€ every 3 years so as long as yhe cameras keep getting better.

I have zero interest in folding phones though. Theyre less durable and have worse cameras so im sitting this fad out.

The article is kinda dumb, and yes this is clearly the area where frontier models having and advantage matters the most, but I'd point out that these smaller open-weight models are performing better than the big Frontier models of just 4-6 months ago.

This means that the Frontier labs are under immense pressure to maintain that lead, and could end up in serious trouble if they stumble at all.

The other thing id point out is that a lot of us who are token-sensitive do things like build plans using expensive, smart models, and then execute those plans using cheaper dumber models.

Then there's the fact that we are still in the age of heavily subsidized Frontier subscriptions + tokenmaxxing initiatives from megacorps. Neither of which are sustainable, and will drive more usage to smaller open models once they end.

LLMs dont do gradient descent to generate tokens.

They are trained by gradient descent, but inference doesnt involve it.

A high quality 1TB NVME SSD costs 164€ on Amazon right now.

For a 10 byte symlink, that costs one billionth of a euro.

This is like worrying about the money you lose when part of a glass of tap water evaporates.

You can save a tool call by just making it a symlink