I feel like modern ML is finding out that you can apply the chain rule to calculate derivatives cost effectively and then scaling up a really really really big

I feel like modern ML is finding out that you can apply the chain rule to calculate derivatives cost effectively and then scaling up a really really really big equation to then iteratively converge on the answer you want.