Serious take: the high turnover of AI engineers / researchers across companies has been absolutely fundamental in keeping information flowing, maintaining competition and the pace of innovation, and avoiding a single-AI provider dystopia.
Meta FAIR's François Fleuret argues that high talent turnover prevents a single-provider AI monopoly
Other researchers credit a lack of non-competes for rapid progress
Positive users praise AI talent turnover for spreading knowledge, boosting competition, and enabling open source via non-compete laws, while negative users see high churn as dystopian with risks of IP theft and power concentration.
No Digg Deeper questions have been answered for this story yet.
Most Activity
Yeah but also don't like steal IP lol There is a thing that you do that generates the insight, that free flow is very important, but like the actual exact thing. Basically just be very very careful
Serious take: the high turnover of AI engineers / researchers across companies has been absolutely fundamental in keeping information flowing, maintaining competition and the pace of innovation, and avoiding a single-AI provider dystopia.
imo if there were non-competes in the AI labs, models would prolly be half as good as they are now.
Serious take: the high turnover of AI engineers / researchers across companies has been absolutely fundamental in keeping information flowing, maintaining competition and the pace of innovation, and avoiding a single-AI provider dystopia.

@francoisfleuret sv was founded on turnover see https://en.wikipedia.org/wiki/Traitorous_eight

@francoisfleuret We should start introducing transfer fees …

@francoisfleuret real

@francoisfleuret Long live open source software systems, the companies that support them, and the universities that use them to teach the next generation of graduates.

@francoisfleuret Anthropic needs to leak some talent, it's getting too concentrated there. Dario seems be having evil thoughts.

@HildeKuehne @francoisfleuret We should start introducing training fees, where companies pay 1 month of salary to the lab that trained the student ;)

@francoisfleuret I just wish they would publish instead: not publishing doesn’t give an advantage against their immediate competitors while publishing is a competitive advantage in retention of researchers

@yacineMTB It's really interesting that AI hasn't figured out the contract stipulation that a lot of other heavily IP-based industries use, where you are contractually obligated to not work for another company competing in the same field for x amount of time (typically 1 year)

@francoisfleuret Maybe but didn't we get turn over from top labs to Chinese start ups and this didn't stop the information flow

@francoisfleuret Result: A monoculture with zero ability to reach AGI, which we should be very thankful for.

@francoisfleuret Which is why any really serious AI research institution must be organized more like a mix of cult and PMC.

@francoisfleuret nope 👎

@francoisfleuret Now tell me, where do Anthropic researchers go to?

@francoisfleuret Now we just need chinese labs to poach talent from openai and anthropic and we're done

@francoisfleuret The high turnover is the dystopia

@francoisfleuret Capitalism winning.

@francoisfleuret Largely possible due to California's law against non competes. One of the few California policies we actually should adopt everywhere.

@francoisfleuret May the hype be with us