With these models, what is the moat? Is it the architecture itself or the training data? Why can’t China create these models themselves? It just feels like theater to restrict models. Can someone tell me why I am wrong if I am?
Its the time and the electricity needed to train them. The US sacrificed quite a lot in order to make this happen. We're tolerating a measles outbreak, our economy is in the trash due to tariffs and draconian immigration enforcement. Basically every sector that's not AI has been poisoned in one way or another in order to drive sufficient investment towards AI.
If China wants to compete at building the best forefront models, they'd have to make similar sacrifices (though I doubt they'd be as messy about it).
But if they can are content with 95% as good at 5% of the cost, then they can focus on distilling instead. I don't think the architects of the sacrifices anticipated just how easy models would be to pirate.
So the most is real, but it's effective for a rather constrained set of scenarios.
> We're tolerating a measles outbreak, our economy is in the trash due to tariffs and draconian immigration enforcement. Basically every sector that's not AI has been poisoned in one way or another in order to drive sufficient investment towards AI.
True, but sustained low interest rates (they are still way too low) cause investment to be moved from productive activities to speculative activities to completely ludicrous activities, as the saner tiers fill up and there's still money left to allocate.
Yeah, I'm not sure if I would call it "incompetence", but it wouldn't be hard for the U.S. public or private sectors to afford MMR vaccine doses for the whole population (or merely the portion that has missed doses on the recommended schedule). Those doses would have a super-low opportunity cost.
MMR vaccines would be a net positive to the US economy, to the tune of $3.5B in direct savings (14x benefit/cost) and $7.6 billion (26x benefit/cost) in wider societal benefits [1]. Those are 2004 figures, so the dollar amount savings would be even bigger now.
To a point made further up the chain: The US's woes aren't driven by AI investment, but the self-interest of a select group. Compare it to a thief who causes $20,000 damage to steal a $100 stereo from a car.
Sure, over decades, but if you think the singularity will arrive sometime next week then all of this forward thinking nonsense is a distraction. And if you're heavily invested in some story related to that singularity, then you have a financial incentive to sabotage the forward thinking nonsense.
If it were my scheme, an incompetent president would be my first purchase.
Sorry this is not true at all. O
Those are just corrupt, incompetent practices by an idiot who is drunk on power. It has nothing to do with AI. This is not an answer at all. You tried to shove a square peg into a round hole for what reason. I don’t know
It's clear that the idiot does favors for the highest bidders. If you follow that money it points you, in various ways, to the AI companies.
Who would have an incentive to tax physical goods entering and exiting the country? Somebody who is competing for investment dollars and whose product is digital.
Who benefits from immigration overreach? People whose products are especially relevant to mass surveillance.
Dismantling the US biomedical superiority and our public health endeavors might've been Trump's idea, but it happened in a room full of people who sell capabilities related to convincing a nation full of sick people that there's nothing untoward happening to them.
The market is controlled by AI companies right now, and Trump is letting the market have its way with us, therefore the AI companies are having their way with us. This is what happens when you abandon antitrust efforts and end up with companies that are too powerful for the government to regulate: they end up running things. It's the robber barron's era all over again. You can't absolve them of responsibility just because the president is OK with it.
And that's what the moat is made of. There will be opportunity costs associated with putting everything you've got into datacenters, and those are costs that most governments are too competent to be tricked into paying.
Yes I wasn’t questioning that the market and government is controlled by the AI companies. What I am saying is that cannot be a moat. If anything it will be an anchor that will pull these companies down. If anything that is a signal that they don’t have a moat
If the scale necessary to get the job done cannot be achieved without a corrupt government in your pocket--somebody to drag people out of town hall meetings and ensure that the infrastructure gets built regardless of consequence--and your competition in another country can't influence their government to do the same thing (and that government itself doesn't care to do so), that functions as a moat, doesn't it?
It's not a new dynamic. The US lets companies do all kinds of things to its citizens re: privacy or the environment or disinformation... things that would be illegal in other countries. It's the American trade-off and although I don't appreciate the explosive diarrhea angle to it, it does give American businesses some protection against foreign competition.
all of the above. the harness is important when you want an agentic process, the model is important when you want tuning. it's a lot like how nuclear research often requires lab equipment, a corpus of knowledge, and access to materials (e.g. isotopes). all of that is restricted in an effort to prevent nuclear proliferation.
If China wants to compete at building the best forefront models, they'd have to make similar sacrifices (though I doubt they'd be as messy about it).
But if they can are content with 95% as good at 5% of the cost, then they can focus on distilling instead. I don't think the architects of the sacrifices anticipated just how easy models would be to pirate.
So the most is real, but it's effective for a rather constrained set of scenarios.