"Films on demand" - some aspects of this are near-future plausible given existing script, mocap and voice acting. Simple programmatic cinematography is just about possible, and AI editing is getting better. 5 years away, maybe, for sitcom / soap-opera equivalent lighting and cinematography. A LOT longer before you're replacing Roger Deakins, though.
"Photo-realistic" - photorealistic CGI films have been just around the corner for 15 years now and continue to be so. Proof-of-concept 15 second renders are doable, feature-length films with non-humans are doable (albeit with a LOT of human intervention), but 90 minutes of CGI humans is a lot harder.
"Characters" - moulding their appearance is almost doable now. Motion is a lot harder - we've got semi-programmatic facial animation but it's a bit rubbish. Programmatic body animation is getting there. Programmatic voice acting is a Really Hard Problem and I'm not aware of anyone making any significant moves forward in that area.
"Unique scripts" - no-one has demonstrated anything close to an AI scriptwriter at this point. It may well be that's a problem which requires strong AI to solve.
We might be looking at Hollywood being replaced completely at some point, but I doubt it'll happen in the next 20 years.
However, what IS a huge threat to Hollywood is the increased power of indie filmmakers with technological assists. I write about that sort of thing over on my blog at http://www.strangecompany.org/blog/
One filmmaker today can do things that would have required a crew of 20 back in 1993. The cost of filmmaking is plummeting. And that certainly is a threat to Hollywood.
