Will Artificial Intelligence Totally Transform Culture & Intellectual Work? An Engineer & Author Debate. Part 1, Music & Movies

Pictured: Barbenheimer meme created through generative AI programs Midjourney and Gen-2 

In the near future, will we keep wanting to watch Hollywood movies like Oppenheimer and Barbie, works of visionary highly personal art by beloved human artists -- or will much of their work be replaced by generative AI programs?

In recent months I've been having a friendly Whatsapp debate on that question with Adam Frisby, CEO of metaverse platform developer Sine Wave Entertainment, one of the smartest technologists I'm lucky to call a pal. Even though Adam takes a highly AI maximalist position on that question!

“Every intellectual job is at risk," as he recently put it to me bluntly. "The ones which can be easily outsourced already today, will be first. Even very prestigious, difficult and stressful jobs like C-suite execs are at risk honestly. Humans can only handle summaries, AI can handle huge quantities of raw data; long term, we're at a disadvantage.”

As longtime readers to my blog might guess, I'm highly skeptical of that -- and not just as a professional writer devoted to the human aspect of technology in the Metaverse and beyond. I've been working in Silicon Valley long enough to have heard grandiose predictions like the above before (and then maybe yet before that before). I tend to think AI will just become one more technology tool among many to produce movies and other art -- but not fundamentally change who makes them. (I.E., human producers, directors, writers, actors, and others talented artists and craftspeople.)

But maybe this time the grandiose predictions are different?

Recent news around OpenAI and Sam Altman recently re-joining as its CEO after a week of confusing tumult may accelerate the issues we discuss, given the org's heavy involvement with Microsoft. Disney, historically a very litigious protector of its content, recently asked Microsoft to remove some its IP out of its AI image program, and may ultimately see Microsoft as a potential competitor. 

Anyway, read our conservation and decide for yourself!

PART 1: How Artificial Intelligence May Change Movies and Music

Wagner James Au: I’m sure you saw that the Screen Actors Guild got the studios to forbid use of AI characters in movies and TV without the underlying actor’s permission. So that puts the kibosh on AI totally transforming popular culture, right?

Adam Frisby: Not at all. “Why pay for a real actor?”. Humans are expensive, notoriously difficult to work with, and the studios are investing vast sums in building someone else’s brand that they don’t even control.

(This is not to be callous, or to devalue artists' contributions to media - just that, as soon as a machine can do a human's job reliably and cost effectively, it will; no matter where or what that job is.)

WJA: But name actors are integral to funding and marketing movies. Even when it’s a CGI animated movie, studios depend heavily on having A list stars doing the voice work.

AF: Yeah; but that's just a matter of branding. AI can be tuned to reproducible results, studios could invest in their own stable of "actors" and probably will. There may be initial backlash -- it'll likely happen somewhere small first, TV maybe? Where novelty is okay.

It could happen via some VTuber initially. That's actually more likely -- low risk, already happening, has a human heavily driving it, someone gets a high profile from Twitch or YouTube, "stars" in something bigger -- low budget film or a TV series, as their character.

That goes on for a little while, the owner licenses the character more fully; precedent has been set.

WJA: But how would that supplant how Hollywood studios work with human actors, when they’re already bound not to use AI actors without permission for 3 years and SAG has no motivation whatsoever to change that in the next contract?

AF: AI versions of existing actors, and permission by the actor themselves. Hell, another route could be someone like Bruce Willis. Someone who for a tragic reason needs "AI help" to act again.

Or, it'll be someone who does something Avant Garde -- "The Film With 10,000 Endings"

However it happens, it'll start small. The studios are not fools. They're aware everyone in the industry is nervous. But that toehold will grow; and eventually the economics of brand name actors will be broken. Because the human actors -- and the supporting infrastructure (cameramen, etc) will be a huge expense for an ever decreasing proportion of the budget

VFX will be long gone by then -- that'll eventually be 5 guys running prompts over raw footage.

Look at this, this is pretty preliminary and early tech, but you can easily see how quickly it'll be able to do complex VFX shots [watch below]:

AF (continued): It’s an economics issue.

Which is better: a small film, $3M budget, brings in $10M or a big film, $300M budget, brings in $700M.

The answer is the small one - it's lower risk. You can run 100 bets for the price of one, if the big one bombs, it hurts, if the small one does, oh well. The reason studios don't invest in primarily smaller films is there's a lot of overhead managing so many projects - smaller AI led teams may solve this.

WJA: That doesn’t factor in marketing the movie, which typically is about the same cost as the production budget - which again is a key reason why studies depend on big name popular HUMAN stars to help promote it to audiences!

AF: I think you're overestimating the value of a human. Hatsune Miku. She's a celebrity in Japan as you know.  "She" I should say. I bet she's already in film, I'd be surprised if not. And look at the popularity of VTubers. There's a human driving those, for now; but that's basically real time voice acting. How long until someone automates that part -- and no one notices.

WJA: I don’t see Miku movies listed in her Wikipedia. Also she was created in 2007 which suggests she’s a one-off exception who didn’t bring in a host of other AI stars.

AF: She's not unique in Japan, there's a few. At some point, AI is going to be as good, if not eventually significantly better, than a human at acting. Just like many, many other industries.

Right now, the only place AI is better is the economics.

WJA: Maybe I missed it but don’t see Hatsune or other AI in this list of top-selling Japanese pop stars. 

AF: That one I don't know -- but just be careful on that since her licensing is a bit weird. She's sold two ways: Anyone can buy a license to her voice and character for a few hundred bucks; and there is an official version who runs concerts and stuff.

And I think a few separate bands have now released under that brand? 

WJA: Yeah I saw one in another list. But again it seems to be a one-off novelty thing.

AF: But if you want a Western example; Gorillaz springs to mind. Not quite the same since the voice isn't artificial. But it could be.

WJA: Same with Gorillaz -- yeah they were big for a while but it didn’t lead to more synthetic pop stars. Also the human artist and musician behind Gorillaz are pretty well known, especially by fans.

AF: Not yet. That's because they're more expensive traditionally. Animators who are good, on top of the cost of the band, versus, the cost of the band.

The only reason I raise these two is to establish the key point: we can appreciate virtual artists, and they can become recognizable brands, which can be used to sell a product.

So a human isn't necessary to make a sale. So then, the argument is down to, "Will AI produce results comparable to a human?", and "Will it make financial sense to fight the established players?"

In the first, technically the jury is still out, but if you compare the progress over the past two years, I'd say it's looking like "yep". Look at the Deep Faked Luke Skywalker in the Mandalorian and Boba Fett, released only a year apart.

In the second question; the cost of these virtual actors is trending towards zero. It won't ever be actually zero, but it'll certainly get close.

WJA: But how can AI completely transform culture if the US and EU refuse to grant copyright and other IP rights to works primarily created with AI (as seems to be where they are leaning now).

AF: It's a very good point. And I don't think the US or EU should grant copyright to it. To be honest; I'm not a fan of the current state of copyright, it was much better in its original incarnation, with sensible maximum terms, and mandatory renewals.

But that aside - trade secret law, NDAs and some kind of human transformation over the final work, will allow someone to add copyright to it. Pride and Prejudice and Zombies is copyrighted despite using copyright free public domain work as the base. 

But the human final touch stuff, is really a loophole - it's a band-aid on the problem of monetising intellectual property. The thing is, as long as it costs money to produce things, there's a need to get a ROI. Creators need to get paid, or we’ll never see new stuff.

WJA: Right so copyright precedent means what will probably happen is AI does the heavy lifting and then a human or team of humans builds on that. And it would have to be more than final touch, to qualify for copyright - it would have to be substantial human input.

AF: No, that's not true at all. It really can be as simple as someone edits the footage, or color grades it. It just needs to be something that touches every frame of the output, in the case of film -- but it doesn't need to be significant. A photographer doesn't need to assemble everything in view for a photo to be eligible for copyright. Their contribution can be entirely trivial.

(Sidenote: the Monkey Selfie Copyright Trial sets the lower bound on novelty for copyright - the bar is not high.)

WJA: Well I guess that could be up to an old school judge who may not even know how to use their iPhone.

So do you think an AI-led Hollywood movie with no humans above the top line (IE star actors / directors / etc) will be made and become a hit? And if so, when?

AF: That's the magic question: it'll probably take a while before one happens with a big budget with lots of marketing. You won't see Tony Stark in an AI produced movie before 2033.

But you might see a Five Nights at Freddie's style indie/low budget hit in the next six years.

The reason I set those timelines is, producing a film, even with AI still takes 3 years to get to screen. The tools aren't quite there yet. But within the next few years they will be; and we'll probably start to see movies made with them within the next five to six years.

WJA: I could see the low budget scenario maybe happening if it is based on pre-existing IP like Freddy’s - which again depends on the humans who made the original IP popular — but don’t think it’ll ever happen with Hollywood movies in our lifetime. But we’ll see!

Read Part 2: How will AI change the Metaverse?

UPDATE, 11:00pm: Updated the paragraph about Altman and OpenAI in light of very recent news.

Please support posts like these by buying Making a Metaverse That Matters!

Comments

[Luther Weymann]

The number one job of a CEO is sales promotion. Growing revenue is totally on the CEO. Falling revenue is on the CEO. The CEO is not a job AI could replace, but it could help with focus. AI is not a replacement for the CEO head of a company meeting the most significant customer and working out a deal to get them to double their orders within the year. Some parts of AI could help, but AI could only be the final decision-maker in niche business. Corporate customers expect human relations to stay a customer, and only the CEO can meet the CEO of the biggest customer for top-level deal-making.

[Adam Frisby]

Luther: I actually agree, I am one myself - but I also am aware of just how many things I am _not_ dealing with. I would say on the average day, I have twenty high priority things I need to do, and have time for maybe three, if I'm lucky.

Summarising and delegating is something I think AIs will do more and more of in the short term - but I have no qualms about saying that eventually, it will replace even that job. The CEO job will become more and more of a figurehead role - as AI eats up more of the decision making in the background.

[Lex4art]

Good interview! The only thing I'm willing to point on is that some jobs feels like "this can't be replaced by AI software" because the way those jobs are traditionally handled (a lot of human to human interaction, live meetings with sales speeches, emotional tricks and so on) but this also may become a subject to change because ... this is obstacle for AI-software. Mean - there will be a competition between companies interacting with each other via traditional human-to-human frameworks and new companies with more and more AI involved and thus with need to develop AI-friendly protocols, exchange formats, data structures. With time very likely it will outgrow traditional ways of doing business interactions - more efficient with the fraction of the cost, fast, reliable... This factor may be so bold that even if we got decent AI software there will be years of slow adoption because AI-powered businesses and organizations need to develop own set of "ISOs" to streamline interaction processes... before that having AI advantage may be like putting very efficient robot on a conveyor, where many other spots still taken by humans and thus this super fancy robot can't speed up things because whole human-oriented conveyor belt can't move faster.

Isolated spots of AI progress is probably the most likely scenario for nearest future (or already happening).

[Argo Nurmo]

@malcolm, chatGPT3 suffers lots of hallucinations. I've used it for research and it suggested (with footnotes!) looking at an existing specialized computer language designed just for the problem I was investigating. This was great, probably a few PHD thesis and academic papers existed too. Wow! Googling it brought up nothing. I searched a couple of academic archives too. Nothing! I asked chatGPT over and over (varying the prompt) and never got a straight answer. Moving to GPT4 I asked about the specific language and finally, after a couple of exchanges it finally said it didn't know of any such language! I hear openAI is working hard to get a handle on hallucinations, GPT4 is an improvement over chatGPT (really just GPT3) but maybe the hallucinations are just getting more subtle and eventually will move beyond were I or anyone else can detect them. -Argo
BTW I sympathize with WJA but agree with AF.