71 karma · joined November 3, 2017
i can't say i support all the advice on this board to find religion. but to each his/her own.
when i was in your shoes, i took some time off, traveled a bit. going to the gym and showering every day is a good plan. Get StayFocusd or similar browser extension, and set a 10-minute total-per-day limit on all news + social media between the hours of 7am and 10pm.
try a musical instrument. it's social, fun, challenging, there's no end to how good you can get if you stick with it, lots of lessons on youtube, and it's relatively cheap (eg - used stratocaster guitar and 20W practice amp, and some cables and a tuner and spare strings and an effects pedal or two and you're set for a long time for <$1K)
If you can afford to be idle, maybe try auditing some classes at your local university this semester. just browse the catalog and pick a few that sound good, nobody's going to notice you in a lecture with more than 50 people in it. computer science, economics, liberal arts, philosophy classes, whatever amuses you. you're learning something and college campuses are full of ideas and energy, and maybe it will inspire you.
But I think the model-based part is still considered relatively hard in real-world environments due to the large search space of possible actions/consequences.
And note that here they didn't actually use the model at runtime, they just used it as an 'expert' (rather than getting a human to provide guidance) to give a model-free agent policy a head start (see DAGGER algorithm) and reduce the number of training samples.
I would have thought that they needed to 'ground' (https://en.wikipedia.org/wiki/Symbol_grounding_problem) at least one shared reference word/concept so they could proceed to define everything relative to that and thereby communicate effectively and know how to map different languages to the same latent space. But I guess perhaps one can infer/approximate it statistically -- https://cis.upenn.edu/~ccb/publications/discriminative-bilin...