@decodyng

machine learning research engineer; lover of cats, languages, and elegant systems; explorer & explainer at https://xitter.cf/t.co/HQSQCz4r5B…

 Oakland, CA
decodyng.com
Joined February 2009
  • Tweets2.8K
  • Following135
  • Followers1.7K
257Media
Looking for a way to catch up on paper reading before NeurIPS? I've just finished a month-long #machinelearning writing project where I summarized papers on everything from lottery tickets to model-based reinforcement learning, and collected them all here docs.google.com/document/d/1…

Machine Learning Writing Collection: November 2019

Context/Introduction Machine learning moves at a bafflingly fast pace these days, and even if some large number of papers on the Arxiv are puffed up beyond their real significance, it’s still the...

docs.google.com
3
74
7
303
Doing the mental calculation of "so, exactly how many papers can I read on the 11hr flight home?" after #ICML2024
2
348
Cody Wild retweeted
Delip Rao
@deliprao29 Dec 2023
Nobody has realized yet that the cat is NVIDIA
Delip Rao
@deliprao29 Dec 2023
AI researcher tuning hyperparameters of their LLM
10
32
339
43,970
Continually feel a vague concern that if all the signage disappeared I'd never escape the labyrinth of the R01-09 rooms area #NeurIPS23
1
257
"Here on the East Coast, we can them LLMs, but I'm told on the West Coast one calls them 'foundation models'? Can you explain?" 🤣#NeurIPS2023
1
278
An O(n) algorithm (including random sketching) to check out the posters is no longer feasible! Does anyone know a sublinear algorithm for retrieval/learning from an unsorted knowledge base? #NeurIPS2023
100s of posters in the poster session

ALT 100s of posters in the poster session

10
5
2
79
24,513
Me trying to find vegetarian food while in New Orleans for #NeurIPS2023
2
1
324
🙏🏼 Grant me the serenity to accept the things I cannot change, courage to change the things I can, wisdom to know the difference, and restraint to avoid getting into fights with people about absolutely any of it on Twitter
44
181
13
1,803
92,102
Christopher Hitchens, in conversation with Elizabeth Wasserman, 2002 theatlantic.com/magazine/arc…
I think Hannah Arendt said that one of the great achievements of Stalinism was to replace all discussion involving arguments and evidence with the question of motive. If someone were to say, for example, that there are many people in the Soviet Union who don't have enough to eat, it might make sense for them to respond, "It's not our fault, it was the weather, a bad harvest or something." Instead it's always, "Why is this person saying this, and why are they saying it in such and such a magazine? It must be that this is part of a plan."

ALT I think Hannah Arendt said that one of the great achievements of Stalinism was to replace all discussion involving arguments and evidence with the question of motive. If someone were to say, for example, that there are many people in the Soviet Union who don't have enough to eat, it might make sense for them to respond, "It's not our fault, it was the weather, a bad harvest or something." Instead it's always, "Why is this person saying this, and why are they saying it in such and such a magazine? It must be that this is part of a plan."

7
43
4
253
30,195
Fairness, social justice, and employment security are not distractions from human existential safety; they are supposed to be part of the solution. Calling these "short term issues" is dismissive and elides their urgency for steering humanity toward a safe and acceptable future.
2
13
49
4,763
What is off-belief learning and how does it help us build agents that coordinate only in grounded ways ? Part 1 of a new blog series on intuitive summaries of key ideas in multi-agent RL: eugenevinitsky.github.io/pos…
2
17
3
66
23,865
Now comes the post-conference minigame of "can I find any vegetarian sit-down food in the New Orleans airport"
2
This is the paper I should have written. #NeurIPS2022 #workboat2022
5
19
3
263
My energy level on conference day 5 vs day 1
2
Appreciating this crepe delivery mechanism at the truck outside #NeurIPS2022
1
Anyone else at #NeurIPS22 a fan of craft beer and interested in checking out @CourtyardBrew tonight? Extra points if you want to nerd out over neat papers while we drink
1
1
2
These are baby stingrays. They look like aliens stuck in ravioli.
851
12,209
1,101
135,853
Every time I read about Classifier Free Guidance for #diffusion, I'm amazed that it works, since it seems like such a weird/tautological hack. Am curious if other #MachineLearning folks have clearer intuitions on it
1
1
10/10, no notes
36
469
60
4,782
We thank the reviewers for their helpful suggestions xitter.cf/FieraIngenieria/status…
35
363
48
3,980
Beginning to feel like #NeurIPS22 poster sessions are a (delightful!) endurance sport that I really should have been training for throughout the year with smaller poster sessions
1
3
Load more