I think part of the problem is that the upper bound on neural nets, as far as we can tell, might very well be general intelligence, and things like self-driving cars, and other nearly magical use-cases that seem within reach. Whereas tree based models, for a series of reasons, many related to scaling, don't offer that feeling of limitless potential.
Maybe. But then again we often try to solve very specific problems, which are very far from requiring anything close to a general intelligence. Heck, a general intelligence might even be bad at things, just like humans are not as good as computers at certain things.
There's a science-fiction story somewhere in there. Imagine an alien planet where brains evolved using a structure completely different than a neural net... some sort of classification model where they were incredibly efficient inside their domain, but broke unpredictably when brought outside of it.
Definitely not my experience in big tech data science. 30-50% of data scientists I work with don’t have a PhD, and took the self-taught hacker path, and earn just as much as PhDs. I probably earned a total of a million bucks by time I was as old as the average PhD data scientist hire out of academia.
To add to this, I do not have a phd in econ, but did what you recommended and work as a data scientist with PhD economists. If you do this, you will make just as much as they do.
Try not to feel this is your one chance and you only have two options. You can get offers again. In addition, you could get work as a machine learning engineer in industry, either now or in a few years. You could spend another year doing deep learning, but dedicate 10% of your time to networking in industry and staying good at interviews, then reevaluate later.
Scientific method is undergoing intense changes in methodology. I suspect in 50 years the way causal inference is done will in many areas be radically different.
Oh absolutely. I often read scientific journal articles which make suggestions of direct relations between different factors when it’s very likely they are correlational.
Example: being overweight increases risk of heart disease.
Alternate Possibility: being inactive/not exercising increases risk of heart disease and being overweight tends to co-occur with not exercising.
In the above example it’s possible that an inactive person who is not overweight has similar liklihood of heart disease.
I quit Amazon a few months ago for the same reason. Holy shit Amazon culture does not work over video. I used to work hard, but I would stay late with my cool team and we’d build things. That stopped happening and I burned out.
They stopped being able to hire seniors easily. So I quit, because working with 8 people 1 year out of college, and myself with 7 years exp, was awful.
After doing 5 years of interviews, I think part of the issue is Amazon doesn’t really do onboarding, so everyone is scared of making an L6 offer, because the new hire is expected to perform equal to an L6 who has already been at Amazon for years, on day 1.
