That AI Temptation
tbray.org
tbray.org
I don't really agree that this is the problem. I think it's much more than human nature interacting with powerful technology creates these "terrible people" and there's no way to not have this "terrible" set of phenotypes expressed when AI exists.
For the record, IMO AI and LLMs are amazing and magical technology. And they raise important existential questions about abilities we once considered hallmarks of conscience and sentience that can be reduced to floating point math. I think people find that threatening having invested quite a bit of ego in being the sole possessors of such abilities. I do not see them as a threat to art: go watch a live performance or visit an art museum. But boy can they create content. And most human content was slop to begin with, see photoslop.
But it's also a cop out to cite that to rationalize the billionaire situation as if everything is fine. It's absolutely 100% not. Musk's investments in Ken Paxton alone should be enough to drive a wealth tax on unrealized capital gains past $1B or so through Congress in 2027, but it likely won't because so many consider themselves temporarily embarrassed paper billionaires. Note that Meta and Zuckerberg have already blinked on this in California. Now if Musk could have just stuck to robots and spaceships, he wouldn't have made himself such a sore thumb, but nooooooo...
And a wealth tax won't solve the national debt. I know that. Duh. But it will start to erode the ridiculous level of influence the top 10 billionaires in tech have right now. The Dictator's Handbook argues that taxation is how leaders transfer money from people they don't like to people they like. Elect people who hate the billionaires and this will all happen organically. Wealth taxes on everyone are bad, that's a given. We're not talking everyone here. We're talking the top 2000 or so richest in America. And if their demographic has no problem going after "welfare queens" I do not see the problem in returning the favor.
Meanwhile, in autocratic China, the AI situation is somehow much healthier than in the states. Go figure.
Is it? It seems to be spreading around everywhere. There are tens of millions of users and hundreds of thousands of organizations on HuggingFace. How are the few of people running the merely most publicity-focused companies in the space in control of the whole ecosystem?
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?
What explains the prevalence of models being trained by small organizations, or even lone individuals, on relatively affordable hardware, then?
> so training a frontier model requires billions of dollars
The term "frontier" here almost makes this a tautology -- "the kinds of models that currently cost billions of dollars to train are only being trained by people who have billions of dollars". But things are moving so fast that the frontier of the immediate moment will be well-trodden ground a few months from now.
Ground-up models are emerging at an accelerating pace, and the FOSS community is producing distillations of proprietary models rapidly.
I appreciate the subjective PoV, but there's not enough engagement with the real possibilities/economics/consequences etc. here to make this a worthy read.