"A PhD student costs the school ~500k" principally because of a set of accounting tricks that universities use to move earmarked research grant money into a general fund, with grad students as a conduit.
PIs experience these as real costs but it is substantially an external imposition by the university rather than a real resource issue.
@Polit_eurOpines Phone bills, utilities, emergency savings in case of health issues? It’s not that harvard is bleeding dry. Most of these institutions CAN pay more. They’re not running out of it… what is this nonsensical bootlicking of uni admin behavior
@ChrisJKennedy@arianaelena97 Bro go back to sucking at geopolitics lol. You’re not even a PhD students or in administration why are you so bothered by this😂
@Hongjia_Lin@TSHamiltonAstro@GregHerczeg Lol good luck retaining a postdoc for a 3-5 year period and get as much value. Postdocs manage PhDs to get projects done. You’re out of touch
@captgouda24@AstroRose20 How is the budget fixed if we're hitting revenues higher than ever before in human history my guy. Leave it to an economist to produce the worst possible ideas with the most non mathematical univariate analysis of problems
@follynomics@AraniLepenque It's not *just* learning. Labs use PhD students to get publications which are the ONLY thing that gets any institution research grants from NSF to Industry. More than 50% goes to the parent institution and towards their revenue. I don't get how hard it is for people to see
People think AI means you can get away with knowing less. But the best performers will be the ones who can give AI the richest feedback signal. In a way, working with agents is RLHF at the problem : propose, critique, redirect and repeat. Grounded intuition is more valuable than ever.
"Even on tasks it has never explicitly been trained to do" suggests that it may have been trained on some tasks? Again, I find it hard to believe that any LLM would have emergent physical AI behaviors with no robot data at all... so there must be *some*
Fable 5.2 shows substantially improved skills in controlling physical robots, even on tasks it has never explicitly been trained to do.
On RoboDojo, a benchmark with a simulated two-arm robot, it outperformed other models and most AI systems built to control robots.
"The long tail is, in some ways, a measure of the complexity of the problem being solved and is directly correlated to the effort needed to tackle it. There are, however, ways to treat it as a first-order concern and build for it" https://t.co/KS7gmj34U8
@Cajalscajones@ChrisJKennedy According to him the only point of creating abundance is to give higher ups more money lol. Let’s go back to pre 1920s and have no weekends either because that’s how it used to be and they should be grateful to have employment