Frontier labs are investing massively in RL environments, yet most of what happens in this space stays behind closed doors.
@chrisbarber and @js_denain interviewed 18 people from RL environment startups, neolabs, and frontier labs. Here's what they found:
I interviewed a post-exit founder friend about how he spends $ to maximize happiness & how he found his vendors
- Finding a cpa, t&e attorney, financial advisor etc
- Finding a chef and cleaner
- How he finds fund managers to invest into
It seems like the highest return on happiness thing he does for life is hosting dinners and events
He hosts one or two dinners per week at his place (4-6 people) and hosts one party every few weeks (30-40 people)
Full interview:
if you haven’t seen the company standard intelligence i’d recommend checking it out (@devanshpandey@yudhister@G413N@_neelr_)
if i was a frontier lab researcher looking for an exciting startup to join i’d look at si
for more info see their blog post about their computer use foundation model, i think this is a good proof-of-work of their specialness
yes, bash style computer use is powerful, but if it’s bottlenecked by things that aren’t accessible to bash computer use, then having coverage for the long tail of software is very valuable
the team shows evidence of outlier ability
will be watching this one
questions i’d have:
- when can their model reach human level for computer use?
- what does reliability look like and does that scale well with compute?
- how do they see bash computer use vs gui computer use evolving?
@Midnight_Captl yeah the leading indicator here could be to try out msft copilot vs cowork on the same set of tasks once a month and then just track when copilot gets good (or maybe it is already)
I spoke about this with @chrisbarber on his show (see clip).
It was clear to me the hyper’s ( like $GOOG) were going to begin to dilute, which is exactly what they’re now doing.
Slight update to my thinking tho from when the pod was recorded - I think they’re going to grow income from ops FASTER than what I previously thought
So this dilution may just be a temporary stop gap to cover this year’s shortfall in net income to pay for the CapEx.
There’s more fuel in the tank than the headlines from this will convey. A lot more. Google is doing the right there here. Now the pressure’s on the other Hyper’s to decide what they want to do (I think they’ll have to follow suit)
@davidmanheim i take it you don't use the 'for you' tab? i think if you sub to a list it just kinda magically increases the weight of those accounts in your for you