Joined April 2018
321 Photos and videos
Pinned Tweet
21 Nov 2022
Our work on improving neural scaling beyond power law won an Outstanding Paper award at @NeurIPSConf 2022!! Come check it out on Wed, Nov 30, at Poster Session 3 in New Orleans.
Our "Beyond Neural Scaling laws" paper got a #NeurIPS22 outstanding paper award! Congrats Ben Sorscher, Robert Geirhos, @sshkhr16 & @arimorcos awards: blog.neurips.cc/2022/11/21/a… paper: arxiv.org/abs/2206.14486 🧵 x.com/SuryaGanguli/status/15…
9
9
109
One of the reasons I’m bearish on neolabs started by ex-Google folks is that they take the velocity provided by the Google tooling and infra for granted See Mistral founder (ex-Meta) interview for CS153 vs Reka founder (ex-Google) blogpost for a reality check comparison
chat is this true?
3
6
206
45,955
carahlo
Alibaba Qwen3.7 slowly fading into irrelevance at the frontier due to proprietary stance. In it's place we have Minimax M3 and... *checks notes* Rio 3.5 397b, made by the municipal IT company of Rio de Janeiro's city government. huggingface.co/prefeitura-ri…
1
19
10,270
Interview question from a neolab mostly made up of PhD researchers: Your colleague suggests we should write our training infra from scratch (a la DualPipe, DeepEP etc) what would you tell your colleague? Me:
1
142
EU doesn't have the building blocks: power and compute
What Europe should do right now: 1. Call all the European researchers working on AI and return them back with same salary (or they can stay but switch career). 2. Fill EU places having GPUs with money, and put those people there. 3. AI partnerships with China India.
130
Don't let Chief Keef hear that youtu.be/uul5wT17m28?si=70RB…
It is as immoral to blow up the ugliest country* in the world as the most beautiful one. * It even applies to New Jersey.
126
Jun 12
Replying to @NYMag
There could never be a more perfect description of New York Magazine
115
they nerfed my boi
2
2
29
6,694
valid question from "garam masala in punjabi chole"
2
11
419
Rookie mistake. I asked ChatGPT to write my first neural network
I wrote my first neural networks in pure C, then in Matlab, then in NumPy, before eventually upgrading to Theano. Since then I have seen and tried pretty much every NN framework ever developed. Some are bad, some are good. The good ones understand API design principles.
4
470
Getting one this weekend
1
139
Thanks everyone for showing up!
Running at full capacity. No idle seats in this room. GPU kernel workshop is liveee
11
274
They hit my boi @yacineMTB with the careers page i'm dead 😭😭😭
180
$$$ AI researchers switching labs recently— * xAI cleaned house and having a hard time refilling talent. Shifted to hiring more startup / engineer grinder types vs researchers. Narrowed focus to code. * Cursor having talent trouble identity crisis: undercapitalized financially relative to team talent level. * Project Prometheus (Bezos) quietly snapping up talent. Many key hires recently. Potential to be major player. * Anthropic remains most desirable, even more than last 6mo, few leave. Difficult to poach from with upcoming IPO. Only hiring staff or above and stopped hiring even senior. * TBD (Meta) also keeps snapping up top talent quietly. MSL seen as significantly less desirable. * Thinking Machines has somewhat stabilized after departures earlier this year. Star studded still but not an auto-pick for talent newly on market. * OAI churning as always, both from normal burnout bleed and latest reshuffle axing non-core divisions. * Not super sure what’s going on with GDM wrt talent. Perpetually #3 spot on model ranking. Overall, talent flow is net flowing out of undercapitalized neolabs to highly capitalized neolabs or Anthropic.
1
302
Made slides and code with Claude and Gemini's help for a kernel writing workshop. Asked Chat to review them harshly. It said the content was r****ded
1
140
*openai-anthropic polycule they take turns being the stay at home parent on their hybrid days
random hypothetical openai-anthropic couple say one parent has to become a stay at home parent which one should quit and stay home
1
190
its...beautiful
Attack on Wemby FINALE
2
277
📢 Two twos my word Torontomans, GPU kernels workshop on Thursday evening is the move fam
A new week begins, and with it a new cohort program. Catalyst runs Monday, Wednesday, and Friday for the next two weeks. Live-coding with PRISM Collective and a hands-on GPU kernels workshop. Tomato presents Will It Grow? on Saturday. Open house Thursday and Friday. Links below.
3
3
7
1,479