The Goat Library

Posted: August 14, 2023 | Author: | Filed under: Artificial Intelligence, Creativity, Technology | Tags: , , , , , | Leave a comment

Talent as a distribution channel, AI and eternal talent

When the world started writing, en-masse, about the explosive arrival of generative AI many journalists made predictions about who would be AI winners and who would be losers. The inevitable fears, that any new valuable and scalable technology will bring mass unemployment, were tempered by the equally abundant observations that such innovation cycles usually see a transfer of jobs to newly required skills, rather than a simple destruction of jobs on an aggregate level.

The transfer argument, while historically accurate, rather glosses over the fact that often the new jobs do not necessarily go to those that have lost their old jobs. The common accusation thrown at those that decry such innovation, that they are Luddites, is in fact quite a distortion of the actual complaints of the Luddites. They most certainly did destroy the new technology as it appeared in ‘their’ factories, but the complaint was not about the arrival of new efficient technology, rather that they were to be cut out of the value it would generate and left with no way to support their families and communities. A fear that played out as they suggested it would.

It is, therefore, reasonable to ask who the losers will be because there will be some.

One sector that got an early mention in much of the coverage is the voice acting industry.

Apple, for example, has already offered AI voice to creators for audio books. It seems inevitable that AI voice will eventually decrease the size of the established voice market that is currently served by human actors.

Sadly there is a bigger problem for the voice community which will inevitably hit other creative sectors too. Eventually, and it will take a while longer, a library of Goats, a library of the greatest voices of all time, will establish itself killing the capacity for new market entrance to all except a smaller number of generational talents.

Think about how you build a career in sport. Maybe you are a gifted 14 year old and you dominate in school, have trials for the regional team and then maybe even the national team, all while you’re still 14, playing for the under 15s. Eventually after a stellar school career you look to the professional game, but now you are competing not just against your own age cadre but all the credible adult players too. Now you are competing against all the kids who were 15, 16, 17 and 18 when you were 14 and everyone else already in professional employment. Depending on the sport that might mean playing against opponents up to age 28 or, as is the case in Formula One today, maybe you are now looking at fighting for your place against 40 year olds as well.

The sporting world has one big advantage over much of the creative world because sport must occur in real time. Sure, you can watch old matches or archival footage of the greats but the live experience is treasured for obvious reasons. Indeed it is the primary value proposition of any contest as entertainment, who will win? The advantage that sport has in this context, is that, eventually, all the Goats will stop playing and new generations will step up to replace them. Even if we could digitally capture and transfer the skills of the greatest, we still hold to the fundamental concept of zero artificial enhancement in sport, which means we would baulk at the idea of implanting today’s youth, for example, with Pele’s footballing chops. That is not the case with our creative industries. If you are a voice actor in the future you will be competing for work against the greatest voices of all recent history, living and dead. Voice will become a tangible post-mortem asset. That wasn’t true until this current technology cycle, but it is now.

A counter argument in favour of the status quo is that the capacity for human voice actors to improvise or to understand and translate the emotional nuance of a piece, and then reflect that in their performance, cannot be replaced by AI. There is something in this but even here there might be a Goat library correction. Such interpretation is currently executed by those that have the vocal characteristics for the piece in question, today’s voice actors. But once those uniquely held vocal skills are replaced by the Goat library then a smaller group of specialists, using their natural voices, will be able to offer their interpretive and improvisational services across a much larger range of work, the actual vocal characteristics for the work will be transplanted onto the architecture created by their natural vocal performance. David Attenborough might still be the voice of nature documentaries well into the next century or more.

There’s no reason to think that a Goat library could only apply to voice acting.

For some time now CGI has been used to replace actors who have passed away either before a sequel is planned (Peter Cushing in Rogue One, 20 years after his death) or as a result of an untimely passing before a movie can be completed (Oliver Reed in Gladiator). There is already a specialist field of work for CGI actors, skilled no doubt but cheaper than the box office stars, living or dead, that they replace.

Google and Universal Music have just announced a partnership to produce tools that allow fans to produce AI generated music but that also enables the musicians to receive payment in recognition of their intellectual property being appropriated, either in their voice or their musical style.

Grimes has offered a partnership agreement to musical creators who want to use her vocals on their music, original or AI produced, as long as she gets a 50% cut of the royalties.

This is a tech continuation of an already established trend in the music business whereby big name’s will only sing on music written by others if they get a share of the song writing credit, because that’s where the money is. The fame and following of established musicians has become a distribution channel in and of itself. Talent itself has become a distribution channel and a powerful one at that.

The impact of social media is a key component of the talent as distribution dynamic

Today’s digital social media dynamics have already skewed the money distribution of creativity, creating a huge power distribution with a smaller number of bigger winners and a long, getting longer, long tail (more artists, less money). Taylor Swift is a massively talented musician but she is also a master of the social domain, both of these talents contribute to her economic reality. It is quite possible that AI will contribute to a deepening of this power distribution, not for artists that are alive but by extending the effect to some of the dead ones too.

Did you write a great song? Can you get Elvis to sing it for you? Tupac? Maybe, and maybe it’s not so far away that this might be the most effective distribution strategy available to you.

The Hollywood actors and writers strike is also about digital distribution via captured talent, even if the headlines are mostly about AI and job loss. The studios are trying to lock in this new economic surplus to their advantage, and, subsequently to the disadvantage of the creative talent.

Like all assets the economic power of creative talent lies partly in it scarcity. Today that economic power is shared between the industry and the talent itself. In a world full of Goat libraries the economic power share will change because the talent itself, not the person, will be tradeable post-mortem, as a tangible asset. An asset not inextricably attached to a living human being with agency. The talent scarcity balance will shift towards a greater abundance allied, quite possibly, with a wider access to that abundance. The industry will have options it never had before. Abundance it never had before.

Years ago I was writing, speculating, about the business model for content on the internet. At the time a much used adage was “Content is King”. I came across an adaptation of that saying. “Content is King, but Distribution is King Kong”. The observation was that ownership of distribution, the theatres, cinemas, TV channels, cable channels and then the internet was more economically powerful than artistic genius. Its obvious once noted but it wasn’t widely acknowledged, and still isn’t. The real takeout though, is that ideally you are both King and Kong, talent and distribution in one package.

If today big artist’s are to become distribution channels then a simple reading suggests that this change should empower them, King and Kong. But this dynamic will eventually, if not immediately, turn in favour of the industry as they alone will be able to afford to build Goat libraries alongside the business architecture required to monetise them.  

There are at least 3 factors working in favour of the commercial side of the entertainment industry.

  1. Social media favours power distribution rewarding fewer big names, often more handsomely, and extending the long tail for everyone else.
  2. AI will extend its capacity to extract the ‘essence’ of an artist’s unique gift. Vocal characteristics are an obvious example but it can and will go further. We can already use generative AI, today, to create images in the style of Caravaggio for example. This has 2 impacts. First the Goat libraries will favour those with a significant canon of work or the time and resource to produce it (training data) and secondly it reduces the scarcity value of an artist’s gift. Your voice may be stunning and unique, but will it outsell Elvis on the same song? Will someone take the gamble when they believe Elvis is a banker?
  3. The scarcity of living talent will be commercially devalued if we can call on the talent of the dead.  

The long tail will get longer and the established big name stars will take more of the money while they live. After they are dead the industry will consolidate its control of both the economic and artistic sides of the business. The incentive for new entrants will diminish, but it won’t disappear, the industry power to gate-keep opportunity enhanced. A significant degradation but not a demolition.

I do try to look for silver linings. I don’t subscribe to the AI doom cults even if I worry about what stupid humans will do with ever more powerful technology. If my logic here turns out to be valid then maybe there can be an upside, even if it’s a small one. Maybe the long tail will inspire/force more localised musical community, a return to the idea that expression is sometimes best consumed as a live performance in small settings. If that sounds ludicrous ask yourself if you would be excited to see your favourite superstar stadium performer in a small club playing to 150 fans? They’ll still play the stadium but much like ABBA they might not be there when they do, widespread musical engagement may become more intimate again?