He notes US citizens are a minority of LLM researchers.
A good time to remind people that in my time doing LLM research I feel like a minority of my colleagues are American citizens.
It would be industry destroying to have to rebuild with segregation for frontier ai research to be legal.
Negative users fear citizenship rules will cripple US AI labs by restricting non-citizen talent and breeding suspicion toward foreigners, while positive users expect gains for Americans through domestic hiring and training.
@natolambert Sounds like a great time for ai labs to start recruiting U.S. talent. Eg, did they try proactively reaching out to all U.S. citizens who scored top 500 on Putnam exam in last 30 years? If no, why not?
@natolambert You know you could have spoken up against Anthropic and all their shady and malicious acts that led to this point rather than saying “props to Anthropic,” right?
@natolambert And: After all the vetting for visas and GC through socials etc — the message is still: suspicious alien
@natolambert Economic gains for American citizens getting AI training (real training, not "learning how to prompt", actual understanding) will be immense.
@menhguin @natolambert Cool. Did they try? If not, why not?
@2021Ufo @natolambert i can assure you, this definitely does not work.
@2021Ufo @natolambert most such people dont want to, and are not particularly skilled for LLM research. maybe like, 10% want to.
@menhguin @natolambert They have high iq, I am sure they can figure it out. As for “10% want to”, that is 10% of 500x30ppl. Did anth try reaching out to every US citizen who scored above 1550 on sat? If not why not? Maybe they would need to — gasp— train them 😱
@natolambert Informal polls at 60%+ non-citizens. ITAR-like controls will wreak havoc on US AI
@natolambert This was my first thought. The order also applies to Anthropic employees. So like...is all internal progress coming to a halt?
@natolambert Is the first time a ban on immigrants in the country was set like this? what a weird hill.
@natolambert That may be the point.
@2021Ufo @natolambert bro. i am telling you. i have literally done this. you are talking out of your ass.
@natolambert the weaponized minimization of the term "segregation" for prioritizing a country's own citizens is crazy
@natolambert Trust the Plan:
@natolambert 3rd AI winter will start if this sticks.
@natolambert the crooks and bigots in DC do NOT care as long as they know what stocks to short in time for their insider trades
@natolambert We should be granting them citizenship!
@natolambert Nothing stops the government from granting citizenship to those who would not do harm and get rid of the rest.
Bad ones should be a minority anyway
@natolambert It’s worth waiting for the dust to settle on this.
It’s possible that the fear on USG side is related to Fable’s hidden output degradation for use cases Ant deemed sensitive.
Unclear as yet whether this rule would apply to OAI or open weight models of similar caliber.