LeCun on deep learning

- Yann LeCun posted commentary reminding audiences that modern deep learning relies on convolutional nets (ConvNets) and vision transformers (ViTs) trained with backpropagation. - He described deep learning as a subset of machine learning that scaled dramatically after internet‑era data growth, tying model success to massive datasets and compute. - The post frames current DL tooling and architectures as incremental evolutions of ConvNet/ViT + backprop paradigms in public discussion. (x.com)