michael tetelman @mtetelman

PhD, AI Researcher at BayzAI. Bayesian Methods for AI, AI Fairness, RL for AD+. SF Bay Area Joined January 2011

Tweets

850
Followers

215
Following

729
Likes

5K

michael tetelman @mtetelman

2 weeks ago

@andrewgwils I think it's already time to compare with "best" human achievements, not majority.

0 0 2 315 0

View Details

@zdeborova @zdeborova , you made a non-trivial statement. Non-memorization means with all other training conditions equal it learns a best approximation of a true distribution, which gives a best generalization. So, you are saying it is not? Do you have an example?

0 0 0 39 0

View Details

michael tetelman @mtetelman

7 months ago

@zdeborova Could you point at studies that analyzes a difference of generalization vs non-memorization?

1 0 1 317 0

View Details

michael tetelman @mtetelman

8 months ago

@andrewgwils DL theory that explains could be useful but what we really need is something more practical, that excludes guessing from finding new optimal architectures.

0 0 0 466 0

View Details

michael tetelman @mtetelman

8 months ago

@hr_unhinged Good job Karen. Do not forget to charge yourself overnight.

0 0 0 5 0

View Details

michael tetelman @mtetelman

11 months ago

Priors are all you need!

Sanjeev Arora @prfsanjeevarora

11 months ago

Insightful blog post from @ShunyuYao12 about what's needed for next level of AI progress. He's been driving AI progress since his work @PrincetonPLI including SWE-Bench. Related but somewhat divergent view from @karpathy who argues (correctly in my opinion) that we need to go

1 5 70 10K 38

0 0 1 81 0

View Details

michael tetelman @mtetelman

12 months ago

@andrewgwils Nah. Until you have a chip in your brain and become a network node you will have to do all that yourself.

0 0 0 165 0

View Details

michael tetelman @mtetelman

12 months ago

@2prime_PKU Could Adam tell how these architectures were optimized vs parameter optimization for fixed architectures?

0 0 1 185 0

View Details

michael tetelman @mtetelman

a year ago

@karpathy Bootstrapping RL with a world model is a way to successful scaling.

0 0 0 38 0

View Details

michael tetelman @mtetelman

a year ago

Cool!

Fenil Doshi @fenildoshi009

a year ago

🧵 What if two images have the same local parts but represent different global shapes purely through part arrangement? Humans can spot the difference instantly! The question is can vision models do the same? 1/15

6 121 599 78K 422

0 0 3 127 0

View Details

michael tetelman @mtetelman

a year ago

@andrewgwils If adjusted for inflation it's actually an axiom.

0 0 0 133 0

View Details

michael tetelman @mtetelman

a year ago

@WKCosmo Ha-ha, here is another real "elegant" formula: abs(e^pi - 20 - pi)<10^-3

0 0 0 87 0

View Details

michael tetelman @mtetelman

a year ago

Transition Matching (TM), a novel discrete-time, continuous-state generative paradigm that unifies and advances both diffusion/flow models and continuous AR generation:

Tanishq Mathew Abraham, Ph.D. @iScienceLuvr

a year ago

Transition Matching: Scalable and Flexible Generative Modeling "This paper introduces Transition Matching (TM), a novel discrete-time, continuous-state generative paradigm that unifies and advances both diffusion/flow models and continuous AR generation. TM decomposes complex

17 46 271 21K 179

0 0 0 146 0

View Details

michael tetelman @mtetelman

a year ago

Integration Flows: a novel generative modeling framework

Tanishq Mathew Abraham, Ph.D. @iScienceLuvr

a year ago

Integration Flow Models "Integration Flow is the first model with a unified structure to estimate ODE-based generative models and the first to show the exact straightness of 1-Rectified Flow without reflow."

3 39 266 17K 169

0 0 1 71 0

View Details

Yoshua Bengio @Yoshua_Bengio

a year ago

The most advanced AI systems will likely exist first within the companies that developed them, with little opportunity to collectively prepare for their impacts. This dynamic creates unique risks highlighted in this recent paper by @apolloaievals and should become a growing priority for AI policy.

Apollo Research @apolloaievals

a year ago

🧵 Today we publish a comprehensive report on "AI Behind Closed Doors: a Primer on The Governance of Internal Deployment". Our report examines a critical blind spot in current governance frameworks: internal deployment.

8 43 200 78K 118

9 65 324 25K 93

View Details

michael tetelman @mtetelman

a year ago

Can we cook neural nets?

1 0 0 87 1

View Details

michael tetelman @mtetelman

a year ago

@andrewgwils Simple models of the world and wish for quick rewards is a road to disaster. That's why principles are more important than models. Principles are wisdom derived from ages of human history. We have to stick to it to make good and make it right.

0 0 0 222 0

View Details

michael tetelman @mtetelman

a year ago

@andrewgwils That may change your future in unpredictable way.

0 0 0 102 0

View Details

michael tetelman @mtetelman

a year ago

@elonmusk Yep. Leave your entropy outside, then enter.

0 0 0 13 0

View Details

michael tetelman @mtetelman

a year ago

@docmilanfar That's so much fun: having a bottle that's never full. But what it means for images? Are there any integral topological invariants on image space?