Discussion
It's pretty simple. You're hearing "We're so scared of the AIs, we need to reign them in" from the very US companies building and selling them for two reasons:
- They're about to IPO, and it's amazing marketing for them
- They want a regulatory moat, since they're deeply worried about the open weight LLMs. They want everyone dependent on this tech and to be *the only ones legally permitted to dispense it*.
@cwebber the funniest thing possible that could happen is they get the regulatory moat, followed by the bubble finally popping and they all go bankrupt
@cwebber Also by a happy coincidence they had this epiphany shortly before an expected rise in interest rates.
@cwebber so, I tend to agree with this pov.
But the current AI safety conversation started when an ex-Anthropic researcher said he was scared, which doesn't fit the corporate astroturfing narrative.
https://www.wsj.com/tech/ai/anthropic-researcher-quits-over-out-of-control-ai-fears-707b7628
@cwebber I mean, maybe the guy was part of a marketing maneuver, but that seems kind of extreme on his part. Or maybe the companies capitalized on the concerns he raised.
@brunogirin @cwebber I mean, almost by definition!
@cwebber Standard issue monopoly throgh regulatory capture. Make it impossible for competition to comply.
They always want a regulatory moat, these fuckers.
As for IPO though - pure speculation here - I wonder if their finances look like absolute muck and they're not quite ready to be seen with their pants down to the extent you need to before an IPO. They could be absolute dogs on paper, and I suspect they are.
Some had it that if IPO didn't happen this year they OpenAI + Anthropic are in trouble. Which would mean they are in trouble, and may be trying to dress it up for acquisition.
@cwebber "It'll be the boxing match of the century! We had the legally register the fighter's hands as deadly weapons!"
@cwebber
Don't worry, everything will be alright.
@cwebber I'm anxious about what happens to the rest of the cloud when the "AI Data Center" bottom falls out.
Unless they've very carefully segmented and air-gapped the LLM portion of the compute, there's going to be giant chunks of the cloud going dark.
@cwebber indeed suppress the valuations then make a killing on the ipo
@cwebber question: open weight LLMs tend to come from China. Would the US regulations reach that far? (pardon my ignorance, in Europe i can use either :P)
@cwebber I have no evidence that refutes--or even casts doubt upon--either hypothesis.
@cwebber Bullseye, especially point number 2.
Corporation == self-serving, full-stop.
@cwebber This feels very war-on-drugs-ish
Install bonfire.cafe
Get the full app experience
@cwebber I'm writing an explainer on this. Should be out in the next day or so. I broke it down into four main parties: OpenAI/SpaceX/Anthropic; capital markets, the US government, and creepy nerds.