ON THE USE OF AI
I do not use AI in any part of the creative process.
That includes the writing, the arranging and orchestrating of music, and every part of recording and producing. It is true of my composition, my prose writing and any other creative endeavour. Anything you commission from me - and anything you don't - is 100% James Lark.
I am not against all AI per se, and if I was still in a job that required me to turn out weekly timetables, I would be open to exploring how AI might make that task simpler (whilst still taking ethical and environmental concerns into consideration). Where creative work is involved, however minor the task, I will not touch it.
The AI industry is on a mission to make its product appear indispensable, issuing statements along the lines that people who say they don’t use AI are lying, or that those who don’t adopt it will be left behind. All clearly sentiments that come from people who have something to sell (either that or they are utterly bereft of imagination), but in the face of such propaganda it’s worth explaining why I do not use it - and why my work is vastly superior without it.
AI can’t do what I do
The creative decisions I make are based on decades of learning, experience, creative evolution, and a heck of a lot of intuition, which allows me to write music that is not only first rate, but also original and unique. Even if AI could write music that matched my level of technical understanding (I haven’t seen it come close), it can’t do the rest.
AI can’t think, it can’t feel, it can’t experience, it can’t form emotional responses. The PR desperately wants you to believe that it can, but it can’t: all it can do is imitate those essentially human instincts. And I don’t want the art I experience to be any kind of imitation.
AI can only create things that work by regurgitating an amalgamation of things that it has been trained on. Which means that AI ‘art’ is always unoriginal - and, at best, the very definition of average.
As to the promise that AI speeds up the creative process: the creativity is the process. To delegate that would be to skip the details that make my work as good as it is – as well as to deny myself the creative journey, and the artistic growth that goes with that. (I wrote in more depth about this in my June newsletter.)
AI is theft
AI has been trained on the wholesale theft of copyright material without the permission or knowledge of creators. The process of destructive scanning effectively also steals work by removing it from the public domain. This plundering of work is ethically indefensible: Microsoft Director of Applied Science Brent Hecht described AI scaping the ‘largest theft of labour in human history’. In spite of which, governments the world over have failed to protect the rights and livelihoods of artists - in fact, the UK government has proposed changes to copyright law that would make it easier to build their products using other people’s work.
I am one of the 50,544 signatories on this Statement on AI Training and one of the artists listed on the album Is This What We Want. For further information and resources, see the Independent Society of Musicians and the Society of Authors.
AI is bad for the planet
This ought to be stating the obvious, but awareness of the scale of AI’s carbon footprint and energy and water consumption remains limited - not least because the environmental disclosure of tech companies is often insufficient to measure it. The impact has been exacerbated by the haste with which data centres have been approved and thrown up, without time given for policies to adapt. What we do know is that AI’s global water footprint could reach 4.2 to 6.6 billion cubic metres by 2027 (equivalent to more than half the UK’s total water usage) and that two-thirds of post-2022 data centres are located in water-stressed regions. As somebody who is currently living through a drought for the second year running, I’m quite keen that the priority for water usage is human beings.
Data centres are far from being the only water guzzling technology. However, AI data centres currently have a rising footprint which is putting strain on a situation that is already at critical: moving forward we need genuine transparency and regulation that will ensure that AI companies are held to the highest standards of sustainability. We have nothing approaching that at the moment.
It’s going to cause a global recession
…I mean, this one’s obvious, right? We’re already seeing the warnings and the financial markets are working as hard as they can to put it off, but when it comes it’s going to hit worse than the tech bubble of 1999-2000. Some would say it was a mistake to concentrate 40% of the economy into the hands of 10 companies, but here we go again.
It's not a reason not to use it, but it's a pretty good reason not to become in any way dependent on any of it.
AI is insufficiently regulated
All of the ethical, environmental and financial issues I have listed stem for a failure to properly ensure that AI companies are properly regulated (or indeed, in some cases, regulated at all); governments around the world have been so keen to jump on the bandwagon that there has been a widespread failure to scrutinise it in the way that would be expected in virtually any other industry (consider what the response would be if, say, the energy sector announced there was a 10% chance their product would destroy humanity in the next ten years…). Tellingly, the UK’s newly-appointed Minister for AI, Kanishka Narayan, only ever pops up to talk about the wonderful opportunities offered by AI, rather than about the ways the government intends to protect citizens from its potential dangers. I know which way I think somebody in his position ought to be serving the country.
We don’t really know where the vast profits of these tech companies are going (or if we do, it’s nowhere nice) and we certainly don’t know what values they stand for (or if we do, it’s nothing nice). This is important because, for all that they really want us to believe that AI systems make their own decisions and develop their own philosophies independently, they are obviously the product of the way they have been designed and prompted. AI did not become racially biased or keen on sexual abuse in a vacuum – and I’ll venture the statement that it didn’t become an existential threat without some help either. AI has to learn from somebody, and sadly, in the absence of firmer control, it appears to have taken many of its initial lessons from dickheads.
AI is a rapidly evolving technology and approaches to it continue to evolve as well, so it is only sensible to keep an open mind. I welcome any feedback and discussion on any of the above.