How do researchers keep up in the area if they're in a hot area like machine learning?

Question

I'm wondering how researchers would keep up with all the new works if they're in a fast moving area like machine learning. There's significantly more research being published, so it would seem to be a major disadvantage to continue working in that area because you need to read so many paper that you wouldn't even have time to think up an idea and work on it.

Answer 1

Frankly, you just accept your limitations at some point and find a healthy pace. It is utterly impossible to keep up with everything. Personally, I have found that having a schedule for reading literature routinely helps: do it, say, after breakfast and then move on unless you need to look up something specific.

In pre-Internet era, people independently discovered ideas all the time. Guess what? They still do. There is no shame in that as long as you put a reasonable amount of effort in. But the "reasonable amount" varies person to person. Find an amount of literature absorbed daily at which you are productive and try to keep it up.

Furthermore, consider doing it in stages: there is some research in well-known venues which helps you stay tuned and connected, skim that. There are some of your own ideas: pursuing them costs time and money, so look them up to see if someone has had the same idea before. It is essentially a cost-efficiency analysis: sometimes you might spend a week chasing something you could read about if you could find it before, and that is fine. After all, you have probably gained something from doing it yourself, after all. Spending a few months the same way? Ouch.

Answer 2

A lot of the work even in a fast-moving area like ML is pretty derivative and incremental. If that's your area of expertise, you can probably scan and dismiss a lot of it pretty quickly as not that special.