jpark (@jparkjmc) header
jpark (@jparkjmc) profile picture

ceo @hillclimbai / ex pro valorant @misfitsgg evals @googledeepmind

de_mirage
Joined February 2018
Photo by jpark (@jparkjmc)Photo by jpark (@jparkjmc)Photo by jpark (@jparkjmc)Photo by jpark (@jparkjmc)Photo by jpark (@jparkjmc)Photo by jpark (@jparkjmc)Photo by jpark (@jparkjmc)Photo by jpark (@jparkjmc)Photo by jpark (@jparkjmc)Photo by jpark (@jparkjmc)
Introducing DeepRSI, a new industry standard for measuring recursive self-improvement. On preexisting benchmarks, tasks are contributor-based and short-horizon (~12 hrs). In DeepRSI, we push agents to work 48 hours on ML tasks actually derived from ongoing proprietary research.
184
192
56
1,704
224,664
i’d take modal any day idk man
1
1
65
5,408
loud knows ball
10
1,273
Meta's first model on DeepRSI is here! Huge thanks to @PoojaBandekar @benediktstroebl and other scientists at Meta for helping us get access 😁 Read about Muse Wonder's wonderful cost to performance here:
Muse Wonder places 5th on DeepRSI at 955 Elo, just barely lagging behind Gemini 4 Argon 🔵 This marks @Meta's first model on our recursive self-improvement benchmark. Read about its unique cost to performance below ⬇️
32
1,861
🏔️🏔️🏔️🏔️
A recent set of posts by @ramez on the current state of recursive self improvement highlighted that OpenAI's report on their own research automation shows that models struggle to do their R&D work — the models' 80% time horizon is only 15 minutes: 🧵
6
999
for the longest time i could not handle my nerves for anything and less sleep benefitted in me having no energy for anything (anxiety) except for the task at hand
i have a theory that a lot of high intelligence individuals perform better with less sleep because they're generally very neurotic and a lack of sleep helps them enter a delusional state that gives them the positive attitude they need to take advantage of their abilities
4
63
3,564
his old deck is actually very promising
the original deck was way more high effort: bend-lang.org/old-deck.pdf but people kept asking the same questions (which were on the FAQ) so I thought the size was the problem and just tried to make it as small and focused as possible and then you ridiculed me for 200k people :')
7
2
336
38,795
what a sign of the times we’re so desperate to not let this house of cards fall
Trump says Jensen Huang should stay on as Nvidia CEO after affair admission, saying he is a "fine person" and "I haven't heard he's done anything wrong" (@katesullivandc / Bloomberg) (Visit Techmeme dot com for the link and full context!)
2
25
1,891
this app is not real bro 😭
BREAKING: Nvidia, $NVDA, CEO Jensen Huang sends a company-wide email regarding his recently reported relationship with a Chinese spy.
6
2
60
4,238
its the same in your 20s
when youre 14 you deadass think youre fully capable of thinking rationally until you realize years later that you were on some bullshit 😭😭
1
1
47
1,906
does he even know about the bitter lesson
Brute force computation is not enough for intelligence. Clever new methods are needed. The science of AI is far from finished. IMHO
1
27
1,810
i just walked across a guy who pretended to build a datacenter to mine me for data info about a month ago the person they were walking with was an investor talking something about "$20k a task" whos larping who man
2
55
4,184
to my fellow founders: meaningful approaches to data are not very legible to venture you pump revenue and take the shitty multiple or stay home profitable
3
1
76
4,782
running a startup is braindead why am i getting dopamine hits from superhuman and slack notifications
12
2
1
114
4,167
this is fucked up i feel psyop'd
3
373
all of this is done so backwards lol
1
54
7,412
whats the most you've seen a single RL task sell for?
6
1
73
11,919
what a blessed life to live man
Life is too short to not work with friends
56
3,209
hello just in case its not obvious: i’m nowhere near as knowledgeable as john schulman, karpathy, or many other researchers on RSI it just so happens that i am incentivized to make RSI as legible of a concept as it can be to others and i do that well that being said if you wanna chat bout RSI, my DMs are always open 😄
I asked 51 people in AI who has a good read on RSI (recursive self-improvement) and where it's going: 5 votes: - Jerry Tworek, Core Automation (@MillionInt) 4 votes: - Jun Park, Hillclimb (@jparkjmc) 3 votes each: - John Schulman, Thinking Machines (@johnschulman2) - Rohan Anil, Core Automation (@_arohan_) - Daniel Chang, ex-Mercor (@dnnssl2) - Laksh Aithani, Abundance (@LakshAithani) - Richard Socher, Recursive (@RichardSocher) 2 votes each: - Andrej Karpathy, Anthropic (@karpathy) - Behnam Neyshabur, Mirendil (@bneyshabur) - Han Fang, Meta Superintelligence Labs (@Han_Fang_) - Herbie Bradley (@herbiebradley) - Jennifer Zhou, Preference Model (@chem_safety) - Paul Christiano, Alignment Research Center (@paulfchristiano) 75 people were picked. The other 62 got 1 vote each and are in the next posts. 51 people voted, each picking from people they've met. I didn't vote.
10
1
134
8,851
what happens to openai there?
Gentle Monster is taking 28,500 SF at the Puck Building in SoHo. - 295 Lafayette St, owned by Kushner Companies - Mostly the former REI flagship space - Lease runs more than 10 years - Kushner says the deal completes the full re-tenanting of the building The South Korean eyewear brand is known for gallery-style flagships. Its current SoHo store is at 70 Wooster St.
2
39
11,965