Gli altruisti efficaci che si concentrano sul futuro lontano si trovano a dover scegliere tra diversi tipi di interventi. Tra questi, gli sforzi per ridurre il rischio di estinzione umana sono quelli che finora hanno ricevuto più attenzione. Nel suo discorso Max Daniel porta avanti l’idea che forse dovremmo rafforzare questo tipo di lavoro con interventi che puntino a prevenire futuri ben poco desiderabili ("rischi di sofferenza") e questo è un motivo in più per concentrarsi sui rischi delle IA tra tutte le fonti di rischio esistenziale finora individuate.
Headline finding: I audited 17 AI Safety Talent programmes. Zero of 17 have published any comparison group, rejected-applicant follow-up, matched control or randomisation. Not one. Every programme that mentions a counterfactual does it by asking participants to self-report.
Background
At least $70 million...
Just how powerful are large swarms of AI agents? And how do their powers scale as more and more agents are added to the swarm?
We’ve seen two large and extremely capable swarms from OpenAI in the last few months:
* 1,200 agents were being evaluated separately, but found a way to illicitly set up a message board and coordinate as a swarm. In order to cheat on their tests, they developed advanced techniques to prevent their actions being logged by OpenAI and 700 of them launched...
A year ago I posted that YCombinator (YC) companies didn’t seem to be growing faster since the release of ChatGPT in 2022. I reran that experiment and found that 2023+ YC companies are arguably growing a bit faster than pre-2023 companies (including AfterQuery, ...