Scott Lilwall is Amii's Scientific Communicator.

Sovereign AI has become a red-hot topic over the course of the last couple years. Usually, when I hear it talked about, it’s in the context of defence, or manufacturing, or misinformation. One place I didn’t expect to hear it, however, was in a discussion on AI in arts, journalism  and culture. 

That’s a question that is pretty important to me, personally. While I work mostly in science writing now, I did spend years in arts and music journalism, not to mention a hobby of writing short stories of very questionable quality. Machine learning and what it might mean for creators is a perpetual topic in my circles.

And we aren’t alone, according to Paul Freeman. He’s a visual artist who was part of a panel that started Day Three of Upper Bound, “AI and Canada’s Cultural Sector: Exploring the Issues and Opportunities.”

Paul says a lot of the other artists he knows are “overwhelmed and hesitant about AI” - both if it is something they should use in their own artistic work, and what kind of disruption it will have on an already unpredictable industry. 

The panel was moderated by Simon Marsden of the Banff Centre for Arts and Creativity, and included Blair McMurren —  Associate Assistant Deputy Minister, Dept of Canadian Heritage — and Anne Casselman, a science journalist and writer.

Together, the panel offered a pretty nuanced view of the place that AI might have in the cultural sector. Much of the discussion centred on the lessons learned from the National Summit on Artificial Intelligence and Culture that was held in Banff a couple of months ago, and brought together researchers, artists, and other parts of the creative community together. 

One of the big themes to come out of the sessions was the idea of trust and transparency. Casselman noted that for journalists, knowing the assumptions and biases that are part of an AI model is necessary to use them in a responsible way. The same goes for artists, Freeman noted. He’s been using AI generative tools in his own visual art, experimenting with making massive, billion-pixel digital images that provide immense scale but also the intricate details, in a way that he says wouldn’t be possible with the support of AI tools. 

That kind of trust and transparency isn’t common, especially with the large generative tools that dominate the space, which are largely locked down, where users only see the outputs, and not much about what is actually going on in the model. But Casselman made a strong case that if the current AI models out there don’t provide the trust and transparency that artists and journalists need, Canada would be in a good position to provide it. 

That also brought up the idea of proper credit and compensation when it comes to training data for generative AI. McMurren was quick to point out that Canada doesn’t have an exception in its copyright protection for AI training and data mining, unlike jurisdictions in the US and EU. 

 Let’s face it, a lot of big LLMs have been trained on datasets collected by hoovering up images, words and music from artists who saw neither a cent nor a crumb of credit. I’ve never had to try to make my rent by writing fiction (probably a blessing for both me and for anyone who can read), but for those who have, it is a struggle. No one makes art because it is an easy, lucrative business. They do it for passion. But passion doesn’t keep you from starving. As Casselman noted, even a minimal disruption to artists' incomes could mean a big loss. Already established artists and musicians might be able to weather disruption to the industry. But every world-famous artist and big-name band started out somewhere, usually by scraping by as they built up their portfolio. If AI makes it even more difficult for early-career artists to find audiences (and revenue), how many will “give up and become a barista, or go to law school?” she asked. 

It was a side of the AI art debate that I don’t think gets enough attention. And while we’re a fair way from any definite solutions to untangle this knot. But it was pretty refreshing to see something in depth that went into this under-discussed area of the debate.