“Mainstream” is not just overly vague, but also a marketing illusion.

One person can create a movie to the standards of the past (not counting the actors).

So, even ignoring the “computer generated actors” part:

As soon as one person can create what marketing-think would call a “mainstream” product, then if there’s still a for-profit industry wanting to make money, they’ll tend to invest the effort of hundreds or thousands of people to make something that seems impossible to make with one person, and market that as “mainstream” to deny competition the market.

As for the “computer generated actors” part, not to mention the “exactly to the taste of the (let’s say) novelist” part:

Never. Or, only when the mainstream discernment of the typical mainstream viewer is so degraded that they can’t appreciate the difference, and that’s not just a matter of technology, because humans can’t be fully reduced to AI.

Of course, much mainstream media writing has gotten pretty darn bad and doesn’t come close to representing actual human behavior anyway. And audiences are getting conditioned to watching computer animations, and accepting really extremely dumb films.

The only way it’s really ever going to be possible is by making “mainstream” mean something low enough that that is possible to fake with computers rehashing human performances.

Oh, and also, “just a novel”? I’d say well-written novels are such a different format that they can, do, and probably always will, remain able to express things that films don’t.

Never.
Simply because of the fact that if AI becomes so advanced, that you can just tell the AI what to make, then you are not making a movie. The AI is making the movie, and you are just an advisor.

Never. If AI ever gets that good we’re all going to be enslaved by our robot masters.

Well, I acknowledge that the above first three responses have good points, but I do not agree with the conclusions: I do not think what I have in mind will be impossible in the future. But perhaps I need to be clearer about what I have in mind. And, in fact, the above first three answers are helpful anyway, as they have made me focus on exactly what kind of process I am imagining.

I would now modify my original description of who might make such a film and say that the creator of the film would be be like an original screenwriter or a novelist,. I acknowledge that AI would create the film initially—or rather the scenes. But I think the way this would work is that the novelist would describe the scenery and action and characters (describe them either in writing or orally to the voice recognizable AI), as well as their character’s voices, as they would ordinarily in writing a novel, but might, even initially, be rather more descriptive than is usual in a novel. Then the AI program—with a more advanced voice-recognition capability, probably, than today—would create the scenes, characters, and voices only generically, to begin with. The novelist or screenwriter (lets just say novelist from here on, to make it shorter to refer to) would then elaborate or refine their description.

The process, conceptually, would be like when a witness describes to a police sketch artist a person involved in a crime that the witness was able to remember. The witness cannot draw in most cases, but when the sketch artist shows them what they have drawn so far, based upon the witnesses initial description, the witness might say that the subject’s mouth should be wider, or teeth were crooked, or hair or skin color was a different shade, or the person had a more narrow face, etc., etc.

The process would be like some people normally write, like I write, in fact, for any purpose in those cases where I ultimately want to make what is read look well, well composed, or make it clearer, for instance. ( I am not doing this much here, by the way: I likely won’t review this that much but mainly to just check for acceptable grammar and fix typos.) This method of writing is called—at least by me—“writing by iteration”. In other words you make a first attempt, you read what you wrote, then you change it. And you repeat this process over and over until you have as close to what you want as you think you can manage You do not have to be that good a writer, as long as you are a good critical reader. (So it’s the opposite to how, for example, Jane Austen was reported to have written—if I remember correctly—for supposedly she wrote whole novels straight off with little need for revision.)

In the case of the novelist converting their words into film, they would look and listen to what the AI generated on the screen, and then they would (probably orally) correct parts, perhaps just one detail at a time. They would then look at what the AI came up with in revision, after their correction—much as the witness to the crime looks at the sketch of the subject that the police artist drew and corrects it.

So, after many many “iterations”, i.e., after many cycles of correcting and presenting, by the novelist and the AI program, respectively, the film, including the scenery, the characters—how they look and sound and move—will start to conform more and more closely to what the novelist has in mind. At the end of the process, what is on the screen will be pretty much exactly what the novelist had—or then has—in mind,

For the novelist does not even have to have fleshed out completely everything in their mind at the beginning: They may decide (once they have viewed what the program produces) even decide that one look, of a character and scene, or the sound of voice, etc., would be better a different way than they might originally have imagined in their mind.

So, I repeat my question—but in particular for some who might have knowledge of the present state of computer technology or CGI, or animation in films, or voice synthesis, or whatever other technologies would be involved—and particularly if you agree with me that it should be possible some day in the future. Because, my interest is not in being told that it is impossible. (I appreciate, though, those opinions.) My interest is in hearing estimates—and the reasons for the particular estimate—from persons who think that this process is in principle possible and who may have the technical knowledge, in one or more of the above technologies, to give an estimate, as to how may years off we are from where this will be done successfully enough so that films as good as todays, and as realistic-looking as todays, could at least theory be made by a novelist.

A final explanation of why I am asking this question: When I watch films or situation series on TV, in the back (and sometimes even closer towards the front) of my mind, I still know that these are actors who are really, in most cases, completely different in character than the characters they are portraying. And the director, casting director, producer, or whoever, who hired these actors after many auditions assessing how each person looked and sounded and how closely they came to what they wanted to portray in the film, had to settle for one of the actors who was available.

Now this process is perfectly fine for film makers who want to make a collaborative production, who want their actors to give their interpretations and for their directors to add their own spin and creativity to the production. I am not knocking that process. It has made very many successful, enjoyable, important, and influential films. But it is not so ideal for a novelist who would like their exact conception to be expressed on the screen,.

For, in the past and in the present day, a novelist who sees their novel converted into a film, has to make the most compromises with the film’s producers, for so many reasons, e.g., to accept what a screenwriter and director think can be made into a marketable film, what the budget is to hire the actors they want, how the script needs to differ (in the producer’s or director’s opinion) from the (let’s suppose even especially successful) novel, etc. But if the novelist could express their story to begin with as a film, by talking to a (sketch-artist-like) AI program that would directly produce the needed CGI and sound/voices, and if they could, by the iteration process I have described, simply keep telling the AI how to change a bit this voice or that appearance or to tweak some action, over and over, through many cycles of iteration, then there would be an entirely renewed profession: the novelist-film maker. And I just happen to think that will happen someday!

But I would like, especially from those technically in the know—but also from anyone with the imagination—to receive some idea of how many years, or decades—or centuries (if that’s really what they think: my guess, as a layperson, is that it’ll be more like a decade or two)—it will take—based upon how advanced the relevant technologies are today, and upon how fast the expert things they might advance over the years

‘So, thanks, in advance, to anyone who would like to try to answer (guesstimate and answer to) this question.

I am a software developer. I have worked on various AI systems for games and simulations. I’m somewhat familiar with voice recognition and synthesis. I’ve done some work with animation and film. I have a degree in literature. I have a pretty good understanding of human psychology. I’ve also watched and appreciated good cinema, and seen too much bad cinema.

I’ve seen, for example, Disney try to present new scenes with not-dead Peter Cushing and 19-year-old Carrie Fisher in the least awful of their Star Wars debacles, Rogue One. In that case, it was CGI over actual human performances, and even that was weird.

It’s not just a matter of adding more technology.

Computers don’t really understand things or think the same way humans do, especially not about art, emotions, personalities, drama, etc.

You can try to encode data structures, or create neural networks that try to symbolically represent those things. But at best you get an automated system that rearranges things provided by humans. For example, computers can generate maps of fantasy worlds. I have one of those in development and it is useful for creating that kind of creative info. But I have to program it with my own ideas for systems of how to generate things. And I have to choose the random details that I like. And I’m just trying to get landscapes I like and save a lot of detailed work. I’m not trying to simulate the behavior, movement, expressions, speech and dramatic interactions of humans in a detailed way.

Now, I do think you could make a system that will do something like what you are talking about. And people are already working on such things. And many of them will share your enthusiasm and overlook the inherent problems in their excitement and/or apathy and ignorance. And, given the terribleness and formulaity of so many Hollywood productions, they may indeed achieve some definition of “mainstream” parity. I’m not particularly looking forward to seeing the results, though.

In particular, thinking of the amount of skill training and talent that good actors and directors have, and the amount of work they do to produce films, I don’t see a computer being better or easier to work with than actual humans, for most purposes. They could be better of course for providing the right physical attributes, or if you happen to exactly want what the recorded performances the computer can provide or modify for you. But it will tend to still be a lot of work to get it to be like what you really want, especially if what you want is not very much like what it already does.

The interface to be able to speak to a computer and have it get your meaning and then apply those changes to a virtual film in a satisfying way, and one that will let you get exactly what you imagined, seems to me to include several unrealistic wishful notions about how that will work out. That interface is currently provided by human animators, who have massive advantages over an AI in that they ARE a human intelligence capable (hopefully, if they have appropriate education) in getting what the director wants. You’re asking for a non-existent high-level AI that also has that kind of skill set.

You WILL find articles that will ignore or be ignorant of these types of issues, and who will tell you this sort of technology is being developed today, and that will be “coming soon”. And there are many cool-looking and perhaps inspiring demos of parts of some such things. But mostly they will tend to only provide some pieces, do the things that they have been set up to do by human developers (not by some AI that really understands the situation), and that are limited in what they can do.

And there no doubt WILL be systems and feature films that have been made by automated systems at some point.

But it’s kind of like how you can get movie music composed for you my Microsoft Music (or whatever it’s called) – it’s just very formulaic random variations on themes informed by human programmers.

So if you don’t care about the quality, then today you can be one person making a film with computer software.

The quality of tools will continue to improve.

But your wish for one person being able to talk to a computer and have that be a way to create exactly what any writer wants a film to be like, and be easier than working with humans… I don’t think that’s really going to happen.

Thanks, Zaku, for answering and providing all that information. It is in fact exactly what I wanted to know regarding the present state of the technology. I also admit total ignorance (up until a few minutes ago when I read your reply) about where the industry is. And I realized even before reading what you have written that we were a long way off. However I still think that someday a ‘novelist’ of great imagination (they often picture exactly how things look, people look, and how they sound, but can only convey a bit on paper, leaving the reader to fill in the details according to their wont) should be able to create ‘on the screen’ (or however images are presented in this future time—e.g., maybe in holographs or ???) what they are imagining, and that an interface analogous to a witness with a police sketch artist (only where the artist would be an AI) could be the way it would work. Of course AI is nowhere near that point yet—I already knew that—but I was wondering how long it might be, and I can see from your answer (and other things I’ve read about) that it will likely be quite a number of years, or decades, from now. But I do think it will happen, and not because machines will ever become ‘sentient’ or conscious—because I doubt they ever will (based upon my own readings in this: And I was trained as a biophysicist, got a Masters from Berkeley in the 70s—but I haven’t done it professionally for years) . But inevitably AI machines must get better at voice recognition, speech synthesis, the ability to interpret the wishes of the user and implement them on screen, etc. I do accept what you say that it’s nowhere near what would be needed technologically. I just don’t agree that a single ‘novelist’ could not create it all—just like they presently do (and have for a couple of thousand or so years done very well, and better all the time)—if they had a good ‘sketch artist’ interface, which I still think someday a machine will be.

I think the form may be possible to generate, or at least select from a range of recordings and parameters. This is already possible to some degree. But it’s copying a very limited subset of what’s possible, even when you allow a lot of modifiers.

What I don’t think will ever be possible, is to generate actual rich new authentic human content with a computer. But many people won’t know or appreciate the difference.

And I think the part that’s not possible is the most interesting part of a dramatic performance.

I appreciate and enjoy rehashes and re-edits and dynamic computer generated content and whatever (in re-edits, games, collages, etc), but it’s always missing a lot of what would be possible with actual humans.

And I fear that the more superficially close it is possible to re-mix and computer-simulate merely the form of dramatic content, the more people will be further disconnected from that most interesting and most human aspect of drama.

My heart even sinks when I see the computer-generated cartoons created for kids these days. Compare that soulless surreal hell to the richness of classic hand-drawn cartoon animations. That’s the type of missing content.

And while the more the technology improves, the less immediately visually obvious the absence of humanity may be, vital elements will still be missing from drama that’s generated by computers rather than humans.