[dilsmatchanov]

I believe you should do what you genuinely find interesting. Go for 1, dig into internals, read some papers, and see how it goes. Even if you decide not to get into ML/AI, learning how stuff works is always rewarding.