> All in all in 20 years we'll all be writing ml ?

If everything goes well, yes. Everything is moving towards static typing with local inference.