Why Do We Automate the Arts Away with AI?

Lately, I have been wondering why we seem to use AI to automate away the few things in life we as humans truly enjoy: the arts. Why have we not used automation to rid ourselves of menial tasks instead?

Can’t someone else do it?

Why am I still loading and unloading the dishwasher every day?

Why do I have to move clothing from the laundry basket to the laundry machine and into the wardrobe again?

Steaming shirts or ironing. Don’t get me started…

Can’t the windows clean themselves? Or better yet, the toilet. And the shower cabin!

Sure, I can hire a maid like a slave time-sharer. But why do I have to pay someone to do such shitty labour at all? Is it because some people derive meaning from menial labour? I doubt the maid agrees.

While there are use cases for machine learning that serve a clear social purpose (e.g. weather forecasting, predictive maintenance, medical diagnoses, protein folding, sorting agriculture produce, disaster response), many are clear cases of attention and money grabbing, especially in (social) media and consumer products and services. Which begs the question: Why are so many uses of AI minimally useful to society?

Hollywood

A perfect example of this is the current rift between SAG-AFTRA, the actors’ union, and the Hollywood studios. The studios want to scan all actors, so they can use AIs to ‘play’ the actors while copy-pasting their appearances, without their consent and payment, of course. The studios want—nay, need—the actors in their movies, but if they can avoid paying the actors, that would really boost revenue. Damn popular actors and their unfair leverage in contract negotiations!

Hollywood is not alone. The music industry is also trying to steal musicians’ voices, so an AI can create music that sounds like the artists without their consent. Musicians already get paid barely anything from labels and streams, but ‘barely anything’ is still more than nothing, so there is room to squeeze.

The end game is obvious: artists prime the AIs, including writers, so that afterwards an AI can write stories tailored to viewers to improve engagement statistics, which another AI acts out using the likeness of actors who have not been informed or paid for their ‘involvement’, yet another AI generates the music for each scene using the sound of real musicians with no one needing any studio time, and at the very end we have the studio executives that rake in the cash.

I am not against technology as a tool for artists to generate ideas or enhance their works. I am against people profiteering off the skills of artists for the enrichment of the few to the detriment of everyone else.

The question comes down to who owns the talents of an actor or musician. Hollywood studios and the music industry have adopted the same stance as OpenAI, Midjourney, Stability AI, and others in the AI space. They do, obviously. If you can make money off foundation models trained on pirated works, surely you can force creative types to hand over their carefully crafted public images and sounds, so the executives can afford another yacht. The regular shmoes (without disabilities, of course) have already been annotating data for AIs for free for years with captchas and their ilk to prove our humanity, only to be out-competed by bots.

Let’s turn that argument upside-down. How would studio executives and record label moguls feel about anyone using their likeness for selling whatever, bypassing their consent as well as bank accounts? I doubt many people people would recognize these tycoons. But how happy do you think they would be to see their faces plastered on billboards selling pills for wobbly weenies, incontinence, or constipation? They’d be on the phone with their team of lawyers faster than you can say ‘parasite’.

You might ask how this is any different from the millions of workers who annotate data for AIs or moderate content day after day only to see their jobs automated away. There are two distinctions. First, a lot of these jobs, while important sources of income, are gloriously dull. And in the particular case of content moderation, it is a job that leaves people scarred for life. Ideally, as a society we rely on automation to reduce the drudgery, so people can strive to be better. More importantly, no one is using someone else’s public persona against their will. With artists, that image is unique, carefully crafted, and the ultimate source of livelihood, their identity even.

Does the end justify the means?

If the product is good, does it really matter where it came from? That question is easy to answer for executives: no. If they can make money ripping off their talent, that’s acceptable, good for shareholders even. The end justifies the means. Even if the end is ultimately a dead end.

For fans, I very much doubt it. I am not interested in an AI-generated Queen album with Freddie Mercury singing again. Or a Billy Wilder noir with long-dead actors brought back from the grave by an algorithm. Who performs in the film, who directs it, or who plays the instruments and sings on a record are vitally important. That they are paid fairly, too. Because it ensures we can continue to enjoy their works in the future. I am sure a lot of music and movie fans think along similar lines.

But what about the masses? That is much harder to say, but I suspect AI-generated content will continue to go viral, which will boost engagement figures, and product managers in turn will misinterpret that as a clear signal from the market that the content is good. It is a short-sighted move that exchanges creativity for engagement, but aligns with corporate incentives. The market will be flooded with auto-generated junk that recommendation algorithms have to sift through, and for which lots of ad clicks can be registered. A victory for the executives and shareholders.

Deepfakes

Exploitation is not innovation. In fact, what I have talked about so far has a proper name: deepfakes. And they have already ruined the lives of women.

Only 12% of leading researchers in ML are female, which is only half of that of computer science as a whole. Note that the percentage of female Kaggle users is in line with these figures, which shows that such platforms are not equalizers, since they attract the same crowd.

So, again, why are we automating away the arts instead of chores? Well, chores are unpaid work, mostly done by women, so why bother? It is much more fun to build an AI that undresses women without their consent and knowledge, manipulates public opinion, or builds autonomous weapons. Not that there are no useful robots at all. Some clear landmines, while others go into nuclear zones.

And we have robotic lawn mowers. That counts, right? Unfortunately, yard work is a task mostly done by men everywhere.

There is a laundry-folding machine and a robot, but they only take care of one aspect of the chore. It might make the overall experience slightly less annoying, but just like a vacuum robot, it requires you to adapt your home and lifestyle to have a ‘robot friendly’ environment.

By 2033, more than a third of unpaid domestic work can be automated. Of course it won’t, except maybe the bits the tech bros with families are responsible for. Men are spending more time on childcare than in the past, which is perhaps why we see an increase in ‘social’ domestic helper robots. Of course, I merely speculate.

The ‘c’ word

I have tiptoed around ‘capitalism’ for two reasons. First, it is an argument of last resort, a cop-out. I doubt any of the viable alternatives would fare any better. Likewise, an ‘algocracy’, a society run by algorithms, will inevitably be corrupted, if it is not already from the get-go because it was designed for someone’s profit, not society. A society made up by a diverse set of individuals with their own preferences, interests, desires, and dislikes, not a homogenous group that mirrors the algorithms’ creators with a set of edge cases to cover everyone else.

Second, it does not absolve us of ethics. We can afford to treat each other better and pay everyone fairly, just look at when it’s time to bail out banks, fund another war, buy stock back, or raise corporate profitability amid fears of yet another recession. We simply don’t.

Yes, money is a powerful motivator and we all need jobs to live, except those fortunate enough to be born into great wealth. But just because there are consequences, does not mean people do not have a choice.

STEM can increase our life span, but it is the humanities that make life worth living: the arts to entertain and provoke, languages with which we can communicate across borders and learn from other cultures, history to understand our humble beginnings and to teach us we are not much more advanced than the ‘savages’ of yore, and philosophy to tell us we can aspire to be better. So, why can’t we use artificial intelligence for social good by freeing us from menial tasks? Why must we use it rob ourselves of the pleasures of music, film, literature, and art by commodifying its creation, devaluing human creativity, and ultimately poisoning the well from which the AIs must drink, merely to increase the next few quarters’ of profits?

Just because we can, does not mean we should.