> Residual connections are more than a trick to help gradients flow. They’re a conservation law.
> Not a hack, not a trick. A principled constraint that makes the architecture work at scale.
back
2 comments
OK, I thought I was reading too much into it but those same sentences also jumped out for me
pangram thinks the whole thing was LLM generated fwiw, as dodgy as AI detectors are it is probably among the best. I don't doubt the author started with their own text, but I think it's been substantially revised via ChatGPT
yes this reads like classic intellectual fellicitatio