Thinking about the AI from google that learnt how to draw… | Artificial intelligence

I was reading this post from a year~ish ago that said that google taught an AI how to draw concepts in a similar manner to how a human does. (https://qz.com/958110/google-took-your-drawings-and-taught-its-ai-to-sketch-on-its-own/)

It got me thinking, we as humans are able not just to picture a representation of a concept in our minds, but also make it do things. If I tell you to think of a person, it pops into your mind, your own concept of a person. Then I could tell you to make that person run, and you could picture how that specific person would move. You could do the same if you would like the person to jump, or fall over, or go under an obstacle, even something that shouldn’t be possible like flying.

Would it be possible for an AI to do this? Say you ask it to paint a man, and it would do so according of it’s own concept. Then say the man has a sword, it would give it to them (based on it’s knowledge, it could put in in his hand, or maybe on a sheath tied to it’s hip) and then ask for it to slash with it. The AI would have to get knowledge of how the human body works, the movements required to wield a sword in a specific manner and would generate an animation of the man slashing with a sword.

You think that’d be possible, and feasible? Would we need Google‘s computing capabilities, or is it still too much for a human to develop it, or computers to run it?