It was announced the other week that James Dean will be returning to cinema screens, 64 years after his death. This should perhaps come as no surprise, given that deepfakes are becoming easier and easier for tech companies to master (see the recent launch of Chinese app Zao), and after Star Wars started experimenting with the technology (spoilers, if you haven’t seen Rogue One) – it was only a matter of time before someone brought an erstwhile actor back from the dead in a major role.
The response from professional actors was quick and unequivocal:
I’m sure he’d be thrilled 😕
This is awful.
Maybe we can get a computer to paint us a new Picasso. Or write a couple tv new John Lennon tunes.
The complete lack of understanding here is shameful.
NOPE. this shouldn’t be a thing.
The thinking behind this is pretty clear, and is hard to argue with: how are up-and-coming actors supposed to make a living if they’re suddenly competing with the established film stars not only of this age, but of every age? Until now, everyone could at least hope that their time in the limelight would come, the path to stardom opening up before them as the previous generation faded into obscurity. Not any more.
The argument isn’t that simple, of course. Dean died when he was just 24, having made only three films. His family are said to be pleased that this is an opportunity for him to make his fourth. Given that his early death left his fans deprived of the future films they felt they deserved, and inflicted arguably decades of lost earnings on his family, who’s to say it’s wrong to use the technology available to try to rectify this?
Valuing the past
There’s an interesting question here. It’s not like we, as a society, are totally against the concept of the artists of the past continuing to draw crowds and even, if they hail from the more recent past, make money (hence the 70-year copyright rule). The David Bowie estate is still income-generating. In classical music, pieces by Bach and co. headline concerts year in, year out, and no one bats an eyelid. Sure, the money in this case may not be going to the Bach estate – but there’s a pretty compelling argument that says that contemporary composers would get more airtime if the focus shifted to the present day.
In any art form – film, music, literature – we clearly, as a society, value that which came before. We respect, and give time to, the classics, be it Sense and Sensibility or The 39 Steps. The creative world is a competition for attention – the human population only has so many hours in their collective days – so, inevitably, any attention given to the creativity of the past leaves less available for the creativity of the present. And we’ve decided, as a society, that this is Ok, at least to an extent. We make some time for the old, and some for the new. So is bringing James Dean back to our screens any different? Is it qualitatively different to people sitting down to watch an old James Dean film – which I imagine no one would mind about?
I think it probably is different. What we’re talking about here isn’t just valuing and reliving the classics – it’s taking a piece of art we’d otherwise expect to be contemporary (with the knock-on economic effects of people working hard today getting paid), and replacing a contemporary producer of that art with a producer from the past. It’s like having an AI trained to compose like Mozart headlining Glastonbury. And this is where it comes back to music.
Former stars, reborn
The technology that would enable the same to happen in music is closer than it seems. What would it take to bring Frank Sinatra back, and have him write a new record, on top of what these filmmakers are going to be doing with James Dean? The answer is that the only thing that’s really missing is the tech to replicate his singing voice. There are other things, like writing new music and lyrics, which will be important parts of this balance of power between contemporary and past artists, and I’ll come onto them later – but all that’s really needed for a new Frank Sinatra record is his voice.
Earlier this month I spent a week at ISMIR, the leading conference in the field of Music Information Retrieval, which is essentially the study and design of computer systems that can understand music. One of the papers presented was on a new system for helping people who have had laryngectomies to sing. It works by taking the speech created using an electrolarynx – an external medical device that produces mechanical vibrations, which the paper reports produces very intelligible but unnatural-sounding speech – and converting this to a singing voice using, among other things, deep neural networks. As the paper concludes,
The results show that the [system] significantly improved the quality of converted singing voices in terms of the naturalness.
This kind of approach is being used to try to help people who’ve had laryngectomies right now, which is an admirable goal. But what’s to stop this same technology being combined with existing speech replication systems (remember this fake Obama speech?) to recreate the singing voices of the pop stars of the past?
Singing voice-replicating technology plus CGI gets us to the stage where Sinatra can come out with a new Christmas album, and even the associated videos. But the songwriters will still be being paid, just as the screenwriters of the James Dean film will. This changes with AI composition technology.
10% of the 110 academic papers presented at ISMIR this year were on the topic of automatic music generation. Everyone I spoke to there said this proportion is growing every year – which is particularly notable given that, if you think about, it’s not totally clear that music generation should fall under the umbrella of ‘music information retrieval’ at all. And this rapidly escalating interest directly reflects the huge leaps forward that are being made in the field.
To put it bluntly, automatic music generation is getting really scarily good. Listen to these examples if you don’t believe me. This isn’t like quantum computing, where there’s lots of hype but we’re potentially decades away from a single useful application. It’s the opposite: most people don’t seem aware of quite how advanced this tech is getting, but it’s already writing music that’s being used all over the place.
What happens when it gets good enough to write a new Frank Sinatra song? Because it will get there. Suddenly, it’s not just the Chris Evans equivalents – the pop stars – competing with musicians of the past. It’s the songwriters, too.
Labour vs. capital
What all of this means is that the balance of power in the music industry could start shifting from labour to capital. Right now, the money goes (at least to an extent) to the musicians working for it: the songwriters, the artists who are recording and touring. Yes, you can keep earning royalties long after you wrote or recorded a song, but there tends to be a pretty heavy weighting in income towards contemporary artists. This is money accruing to the providers of labour. But when artists from the past release new albums, it’s the asset that is valuable – the person who holds the rights to the artist’s likeness, the artist’s voice, is the beneficiary of the income. This is money accruing to the owners of capital. And it means it’s that much harder to make a living as a musician by working hard.
This is no different to fears surrounding other industries. When self-driving cars are everywhere, the money in the taxi industry will flow into whichever tech company owns the cars (the capital), instead of going to drivers (the labour). When self-service checkouts replace the last cashier, it’s the self-service checkout company that will own the capital, at the expense of the income streams of the people it. displaces. But it’s perhaps more surprising in the world of music, where our gut feeling is that artists are safe from automation.
Of course, it might all be Ok, because perhaps no one really wants a new Frank Sinatra album anyway. But what if this revamped Sinatra changed it up and released something a bit more pleasing to modern ears?
The makers of the new James Dean film say they searched high and low for the right actor for the role, but in the end there was no one better suited to it than Dean. This seems to have been corroborated by Tom Brittney, an actor who has said he went for the role and was rejected. This might be the first instance of a film star’s job being replaced by this kind of technology. Who will be the first musician to say the same? And how comfortable are we that that is the future we’re racing towards? My hunch is that we, like Chris Evans and Elijah Wood, might end up not being that comfortable with it.
This was originally sent out as part of a newsletter I write. You can sign up here.