hardmaru

@hardmaru · 27 mai 2026

For over a decade, we’ve accepted that end-to-end backprop is the only way to train deep networks. But holding the entire network in memory all at once is why AI training is hitting a resource wall. We found a new way to break the network into blocks and train them

Sakana AI@SakanaAILabs· 27 mai 2026

Introducing DiffusionBlocks: Block-wise Neural Network Training via Diffusion Interpretation What if we didn’t have to hold an entire neural network in memory to train it? Standard neural net training optimizes all parameters jointly. As a result, the

Elon Musk

@elonmusk

Interesting

19:59 · 28 mai 2026 · 258,5 k vues

166

117

1,5 k

X →