It's my hot take that the next big ML breakthrough needs s.th. that learns from its own actions in an environment, so this goes in the right direction imo.

On the other hand a lot of big companies struggle with self driving cars even though they predicted to build this years ago. Also probably all big AI companies work on AI for autonomous robots. Where do you intend to do s.th. different to get a shot at competing with them (when they have so much more capital)?