One day, when I was trying to stay awake during my economics lecture, I saw how we currently estimated future GDP using something as mind-bogglingly simple as linear regression. Knowing a bit of ML I decided to see if I could use a neural network to improve our estimation skill. So far I haven't. The bias is way too high and it has not-insignificant variance too. I've a few things yet to try but this thing's goal will probably pivot soon. Obviously linear regression isn't great. I'm just not sure how to beat it.
In the meantime, at least I was able to use the economics data array and an image library to make some sort of picture from the last few years' economic data.
If I cannot yet contribute to the economic community, at least I can contribute to the artistic one. In a very, very small, crap way.