Congratulations to my peer @TheOfficialACM Fellows 2022! It is an honor to be part of this group!
@animaanandkumar
-
ACM Fellow Recognition Celebrates AI Research Excellence
By
–
What an amazing evening at @TheOfficialACM awards reception yesterday! Honored to be selected as an ACM Fellow and privileged to be working with phenomenal students and colleagues! Wonderful to have @bjenik by my side and cheering me!
-
GPT-4 Generative Modeling Advances Reinforcement Learning in Minecraft
By
–
Agree. Our work showed that using GPT4 in an interactive way to generate automatic curriculum and build skill library can tackle complex long-horizon tasks in Minecraft. The future of RL will be generative modeling + online learning.
-
Generative AI Overtaking Reinforcement Learning in AI Development
By
–
Generative AI eating RL's lunch. This is just the beginning https://
x.com/AnimaAnandkuma
r/status/1665048562852564992?s=20
… -
Neuro-Symbolic AI: Integrating RL with Generative Models
By
–
Thank you @swarat I agree that neuro-symbolic AI is the future. RL will play some role, but it will not be RL from scratch. It will need to integrate with generative models.
-
Guggenheim Fellows Reception Celebrates Diverse AI and Science Awardees
By
–
Wonderful reception for @GuggFellows in NYC. Awardees from diverse backgrounds including scientists, artists, writers, historians, choreographers. Met @NarangLab in person and Ronitt Rubinfield (who was a wonderful mentor when I was at @MITEECS ) https://
gf.org/announcements/ -
Neural Operators for PDE and SDE: Diffusion Models COVID-19
By
–
We demonstrated it here in Covid-19 research. https://
biorxiv.org/content/10.110
1/2021.10.09.463779v1
… Regarding PDE vs SDE, if you input noise as a variable @CristopherSalvi has FNO for SDE. If you want noise to be captured implicitly, you can use our diffusion neural operator https://
arxiv.org/abs/2302.07400 -
Neural Operators Enable Super-Resolution Across Space and Time
By
–
Neural operators can do different resolutions in space or time, the framework is general. In our first FNO paper, we showed super-resolution in both space and time https://
arxiv.org/pdf/2010.08895
.pdf
… -
Dynamic ratio adjustment balances synthetic data fairness improvements
By
–
Synthetic data to augment under-represented classes in real datasets is natural to improve fairness. However, we have to balance distribution shift with improvement of fairness. We propose dynamic ratio adjustment of real vs synthetic data. @deanh_tw @wn8_nie @ArashVahdat
-
Neural Operators: Discretization Invariance for Variable Resolutions
By
–
Interesting @smnlssn I encourage you to explore neural operators since they will provide discretization invariance and can handle different resolutions directly.