Perhaps a better place to find algorithmic details is this related paper, also with Hinton as a co-author, which implements similar ideas in more standard netwo

Perhaps a better place to find algorithmic details is this related paper, also with Hinton as a co-author, which implements similar ideas in more standard networks: Scaling Forward Gradient With Local Losses Mengye Ren, Simon Kornblith, Renjie Liao, Geoffrey Hinton https://arxiv.org/abs/2210.03310 https://arxiv.org/abs/2210.03310 and has code: https://github.com/google-research/google-research/tree/master/local_forward_gradient https://github.com/google-research/google-research/tree/mast...