Statistically Proving That 'Pre-training Works'—Is ImageNet Knowledge Really Useful in Medical AI?

When you hear that 'fine-tuning a ResNet pre-trained on ImageNet improved accuracy by 20%,' have you ever determined whether ...
aiweekly

QuadTok Cuts ~10% of Image Tokens, Hits 2.08 gFID on ImageNet

A new paper from Zhuowen Tu's group proposes replacing the fixed 256-token grid used in autoregressive image models with a hierarchical quadtree.

Why You Should Avoid Using Models Pre-trained on ImageNet

It is not that pre-trained models are bad, but rather that you are failing to detect distribution shifts. Have you ever ...
HPCwire

AMD Lands Fei-Fei Li, the ‘Godmother of AI,’ In $8.2B World Labs Buy

Nvidia may be buying Hugging Face, but AMD will soon have an AI software company of its own following the chipmaker’s announcement this week that it’s acquiring World Labs, the spatial AI company ...
Tech Times

AMD Buys World Labs: Fei-Fei Li Will Now Shape the Chips That Run Physical AI

AMD acquires World Labs for $8.2 billion in an all-stock deal, naming AI pioneer Fei-Fei Li as chief scientist. The move gives AMD in-house frontier AI research to shape its chip roadmaps around world ...

AMD will acquire Fei-Fei Li’s World Labs for $8.2B

The acquisition will see World Labs founder Fei-Fei Li join AMD as executive vice president and chief scientist.
quantumzeitgeist

Researchers Achieve 78.3% ImageNet Accuracy Using Quantum-Inspired Transformers

Achieving 78.3% ImageNet-1K top-1 accuracy on the ImageNet-1K benchmark, a new vision Transformer architecture called QiT has been developed. The system introduces QiT, a Quantum-inspired Transformer, ...