I think there is an enormous survivorship bias here. Research advances come in alternating of periods torrents and droughts, not in a steady trickle. Hamming, Shannon, and their contemporaries worked at a time of revolution in computer science.

Let's put deep learning aside for a second. Can anyone here name a single academic in any other subfield of computer science who had a string of groundbreaking research works where every one of those works is from 1980 or after? I can't. And that's almost half a century that we're working with. Many people who defended their PhDs in the 1980s are already retired.

This applies to plenty of other fields as well. Medicine, for instance: There were surgeons in the 1960s and 1970s who have half a dozen new surgical techniques to their name. (If you want some examples, look up Michael DeBakey, Mark Coventry, and Thomas Starzl.) You will not find a single academic surgeon today who comes even close to that, and at least in my opinion it's not because modern surgeons lack drive or creativity or the ambition to tackle big problems.

Hamming is probably right that the role of luck in the sense of serendipity is overblown, but he ignores role of luck in the sense of structural factors that are out of your control. Because he and his colleagues largely operated in a context where all those structural factors were completely optimized.

If you decide to start a PhD in a field 10 years before that field is destined to hit a multi-decade rut, if your advisor fails to make tenure partway through your PhD and your lab is shut down, if you get your first academic appointment in an environment where research funding is trending down and regulatory burdens are trending up, these are all factors that are by and large out of your control that can significantly alter the trajectory of your career.