Ask HN: How to get excited about deep learning
3 comments
> I did use various deep learning methods as a black box for particular tasks they seemed good at. However, I never really liked doing it and at this point I'm starting to feel like a dinosaur not being able or willing to adapt to the new reality.
I think this is actually the new reality. Only few people will work on (advanced) deep learning model and the users will only adjust them to their use cases and applications.
I do understand your issue though, because I have been feeling the same about deep learning and never really had much of an application in my professional life either. It just takes too much time to get to a level where you can actually generate insights.
Besides, most people who do deep learning that I know are the first to tell you, to stick to traditional ML techniques until it's not enough anymore.
I think this is actually the new reality. Only few people will work on (advanced) deep learning model and the users will only adjust them to their use cases and applications.
I do understand your issue though, because I have been feeling the same about deep learning and never really had much of an application in my professional life either. It just takes too much time to get to a level where you can actually generate insights.
Besides, most people who do deep learning that I know are the first to tell you, to stick to traditional ML techniques until it's not enough anymore.
Thanks for the answer. On one side I'm happy it mirrors my understanding, but on the other I kinda feel some sadness for simple small beautiful solutions. A lot of models feel like brute force to me.
It's just a hype topic at the moment. Everyone wants to do AI with crazy deep learning models. I am sure, this will subside eventually.
However, I did my PhD right before the (2nd) deep learning revolution and have never understood how people do research in this new field. Most of the things I did and still do are very well defined and rooted in math, so I mostly know if the solution is good after the fact. I'm also keen on implementing these solutions well and mostly from scratch following similar principles there too.
Nowadays there is a lot of stuff coming out that is possible solely due to deep learning. My issue, however, with performing research in the field of deep learning has always been that I don't understand how to measure success. I might more or less randomly get to a system that works and outperforms the state of the art by a small margin but I would still not know if I found something fundamental or was just lucky. The only way to know seems to be purely empirical. Add to this the energy cost of GPUs and the discrepancy of training capacity between individuals, universities and companies and you will probably get more or less a full picture of my perception of the matter.
I did use various deep learning methods as a black box for particular tasks they seemed good at. However, I never really liked doing it and at this point I'm starting to feel like a dinosaur not being able or willing to adapt to the new reality.
What's your take on this? Anybody went through similar issues? How do I get excited about deep learning? Or is it fine to remain skeptical?