Great point. We are seeing more and more than smaller models with better objectives or data can beat big ones! My main point is that an approach shouldnt go away as models get better. Scale is just one way of getting better
RESEARCH
-
Reverse Engineering Systems: Breaking Down to Rebuild Better
By
–
Breaking down already built projects, and systems into smaller components and learning them individually to understand how it was engineered then redesign and build it again
-
Break Down Projects Into Components Learn Principles Rebuild
By
–
True, break the projects into smaller components, learn each components, understand the principles and build back again
-
Paper Review Assessment: Strengths and Weaknesses Analysis
By
–
Correct, some is cut off. The reviewers also give a brief summary of the paper and list two strengths and weaknesses. I don't know the area/work, but at a glance, certainly not the worst review I've ever seen. (Not linking here to avoid a pile-on)
-
Marie Curie: Breaking Barriers in Science 100 Years Ago
By
–
17 Nobel Prizes on this iconic picture in 1927 at the Fifth Solvay Conference! Notice that the only woman in this group is Marie Curie. 100 years ago, imagine all the obstacles and biases she had to go through… And you, who is your idol in science? #innovation #tech #science
-
Deep Learning’s Future: A 2009 NeurIPS Prediction Recalled
By
–
13 years ago at Neurips 2009, 3 years before AlexNet, a young student @ilyasut explaining to my dinner table why Deep Learning is gonna be the future.
Looking forward to #NeurIPS2022. -
Paper Review Quality Assessment in AI Research
By
–
Correct, some is cut off. The reviewers also give a brief summary of the paper and list two strengths and weaknesses. I don't know the area/work, but at a glance, certainly not the worst review I've ever seen. (Not linking here to avoid a pile-on)
-
Overfitting in AI Research: Need Better Global Standards
By
–
2) During lockdown my team and I explored a lot of the GitHubs from China based publishers & found quite a few claiming major breakthroughs to be overfitted. To be fair it also happens with some US researchers. AI needs better research standards globally. Don’t believe the hype!
-
ICLR Code of Conduct and Organizer Authority Questioned
By
–
The ICLR code of conduct (
https://
iclr.cc/public/CodeOfC
onduct
…) states "conference organizers may take any action they deem appropriate." So I guess the organizers deemed it appropriate. -
US dominance in AI innovation: transformers, GANs, and key breakthroughs
By
–
1) most of the innovation in AI is still occurring via the US or US related entities (eg DeepMind). The key defining techniques such as Transformers with Self Attention, GANs, Federated Learning, Deep Reinforcement Learning, Neurosymbolic AI all emerged in the West.