I spoke to @latimes on #ChatGPT on how it can confidently give wrong but plausible answers to puzzles, and it is not getting online feedback on if the text it is generating is true.
@animaanandkumar
-
FNO Integration into Diffusion Models Research
By
–
Indeed, this is very interesting. We have recently incorporated FNO into diffusion models https://
arxiv.org/abs/2211.13449 -
DeepONets Modification to Function-to-Function Neural Operators
By
–
DeepONets were originally defined on a fixed input grid and hence, do not fall under operators where both inputs and outputs are functions. We can convert it into a neural operator through some modifications. See proposition 5 in https://
arxiv.org/pdf/2108.08481
.pdf
… -
Neural Operators for PDEs: Resolution-Independent Function Space Learning
By
–
To solve PDEs, you can learn at a fixed resolution (e.g. with U-net) and interpolate. But error cannot be arbitrarily small this way. Our neural operators are guaranteed to approximate arbitrarily well at any resolution since they learn in function space https://
arxiv.org/abs/2108.08481 -
NVIDIA Internship: Computer Vision and AI for Autonomous Driving
By
–
Come intern with us at @nvidia We collaborate with @ALVAREZ_JOSEM team to build full-stack computer vision and AI models with applications in autonomous driving,e.g https://
arxiv.org/abs/2204.12451 @ZhidingYu @voidrank -
Zero-shot Super-resolution Tests for Neural Operators Benchmark
By
–
But it is crucial to add zero-shot super-resolution tests. Neural operators can output at any resolution since they learn mappings between function spaces, but U-Nets cannot. Can u please add these tests to the benchmark? @rejuvyesh @jo_brandstetter
-
PDEArena Benchmarking and FNO Architectural Improvements in PyTorch
By
–
I agree. Great to see PDEArena as a convenient benchmarking environment for PDE in @pytorchlightnin Also great to see further architectural improvements to FNO by integrating it within U-Net like structure. Would be great if PINO can also be added https://
arxiv.org/abs/2111.03794 -
AI filters and body dysmorphia risks for teenage girls
By
–
I also worry that it can make body dysmorphia worse, especially in teenage girls who are so vulnerable. Being presented images of themselves in Barbie-proportioned bodies and hyper-sexualized cannot be healthy
-
Anima Anand Kumar shares a link on Twitter
By
–
https://
x.com/AnimaAnandkuma
r/status/1597054443937484800?s=20&t=S7fh-Fu6Yompfo_wnH3akA
… -
Generative Models and Emergent Concepts in Latent Space
By
–
I agree with you that the distribution for generation can be on a latent space rather than on the observed space. A generative model can lead to the emergence of concepts without requiring an external target