Both the US and UK governments are considering the use of copyrighted material from the creative industries in training data for AIs, particularly Generative AI. The tech companies want to do this with few restrictions and without cost, citing “fair use”. As it stands at the moment, the UK Government looks to be leaning towards the tech companies but judges in the US are critical of the tech companies’ arguments. Even the EU, which normally stands up to the excesses of the tech firms, seems to be struggling.
As you can imagine, many authors and musicians aren’t happy about this and to be honest, I’m with the artists on this one. I do accept that often art builds upon the work of previous artists but most artists are generous with credit and rarely does anyone seek to pass off someone else’s work as their own. In the music industry, when an artist does re-use an element of another artist’s work, royalties are paid to the original creator and this is perhaps a model to consider.
The AI companies are attempting to avoid paying the creators but I imagine that they’ll inevitably have to do deals with music labels and publishers, but as an independent artist the difficulty is going to be proving that your work has been subsumed into the AI, and that’s why I agree with the efforts to force AI companies to disclose the training datasets. It’s far too opaque right now to know what’s inside the AI. Artists would then be able to claim any of their creations that had been used in the dataset for reward (or removal from the data set).
And let’s not pretend that the AI firms are doing this for the benefit of society. They’re doing it to make money (Google’s AI Ultra is $250 per month) and these are already some of the most valuable companies on the planet. It’s worth remembering too that technology companies have a pretty poor track record when it comes to privacy and other rights. They don’t have anyone’s interests at heart but their own. No-one needs “music in the style of The Rolling Stones”, “stories in the style of John Grisham or “pictures in the style of Basquiat”. It’s not like the AI is reviewing MRIs for cancerous cells or working for the greater good.
Ultimately, the companies behind generative AI need to be respectful of copyright holders. They need to both pay creators appropriately or honour those who don’t want their art included. Please do the right thing for once.