Opinion
I must have been the only person looking forward to AI, until I met Dr Whittle
Malcolm Knox
Journalist, author and columnistWhen I hear the words artificial intelligence, I reach for my hot water bottle. Too complex, too science fiction … too tired … Leave this one to the next generation.
Inconveniently, like climate change or Vladimir Putin or global pandemics or a leopard in the jungle, AI is not letting us outrun it. It’s upon us. So, like a coward dragged to a creature feature, I attended a webinar given by Dr Jon Whittle, director of the CSIRO’s research into digital technologies. His subject was “AI: A point of inflection for the arts?” and his answer was yes indeedy.
Illustration: John Shakespeare. Credit:
Before continuing, I want to stress that Dr Whittle seemed a very nice man, a good communicator with a brain the size of a planet open to all possibilities. He hosts a CSIRO podcast, Everyday AI, to demystify the science for a lay audience.
He is, of course, an “AI positivist”, and I entered his webinar as an AI Pollyanna, when it comes to writing at least, seeing (hoping?) the technology as a very very fancy version of predictive text. I like predictive text. It saves energy, just as auto-correct fixes fat fingers. For writers who have survived the typewriter and the word processor, AI could offer improvements.
Knowledge workers fear AI might make them obsolete, but I thought, if AI can save humans from the soul-sapping work of producing the infinite clutter of formulaic professional documents that seem to be written by machines anyway, leaving the human as an editor rather than a generator of words, that can’t be a terrible thing. And why should writers be spared from the grand project of efficiency and scientific management that has taken factory and manual jobs for the past 250 years?
Unintentionally, Dr Whittle revealed how naïve I was and how worried I should feel. Dr Whittle believes that “AI outperforms humans on many tasks, including any task that relies on assembling patterns and information gleaned from large datasets. Humans are sluggishly slow in comparison, and have less than a fraction of AI’s memory.” On the other hand, humans outperform AI in “commonsense and logic”.
The first alarm bell rang when Dr Whittle talked about AI-generated journalism. It is faster and more efficient than humans. It does get all the facts wrong, he said, as if that was a glitch rather than a fairly fundamental baseline. Pollyanna began to waver. Facts can be either correct or incorrect, but – as we see every day in our world of alternative facts, alternatives to those alternatives and alternatives to those too, everything lies in the “commonsense and logic” – the art – of interpretation and arrangement.
But AI promoters think it can do art. Dr Whittle moved onto music, playing the AI-generated theme music for his podcast. He loved it. Now, maybe I don’t know what’s music and what isn’t, and to some degree this has to be a subjective assessment, but let’s say it was music that suffered from only one flaw, which was to be unlistenable.
This is not the issue, however. AI is not proposing a new musical taste; it is proposing that musical composition is simply labour that obeys “assembling patterns and information gleaned from large datasets”. If that’s all it is, then the act of musical composition is just another task that humans can do without, and therefore (according to AI-hungry corporations) that humans can do without paying for.
With unwitting enthusiasm for AI, Dr Whittle developed his increasingly sinister-sounding theme by talking about visual art. He showed an AI-made exhibit displayed at the National Gallery of Victoria. He had photographed his young daughters dancing in front of it.
Again, taste for the artwork is the least relevant consideration. Maybe beauty is in the eye of the beholder. But machine-produced visual art has been around for a while, at least since Andy Warhol’s screen prints asked us to consider what art is. As a novelty, it asks good questions. But for it to be presented, 50 years later, by scientists as a labour-saving alternative is a little off the pace, at the least. One of the essentials of human art is that it is made by humans, whose motives and puzzles are what open the brain to surprise, mutation and advance.
But humans still have their place, according to the CSIRO: as AI’s artistic “collaborators”. For example, he said, his daughters were “collaborating” with the AI artwork by dancing in front of it.
Fifty years ago (again), interesting questions were raised about the audience’s role in completing a work of art. On the other hand, girls can dance in front of a fountain or a waterfall or a tree. It doesn’t make them collaborative artists and it doesn’t turn a tree into art. There is an infinite depth to artistic production that – at risk of sounding religious – is unknowable, precisely because it emanates from mercurial, mixed up, inefficient humans.
If this is making AI sound incapable of creating anything new, isn’t that its own boast – that it can only plagiarise the work of others? Not in a tricky, provocative, post-modern way, not as recombinant collage, but as plain theft? The purpose of predictive text, after all, is to anticipate what we already knew. Its efficiency lies in saving us the trouble of thinking. Its virtues lie entirely in efficiency and never in originality (unless you are a true academic cynic who thinks “originality” is an alternative fact).
With so little experience or insight around creative processes, then, you wouldn’t think digital technicians would have much say in how the federal government creates rules for artistic intellectual property when it is affected by AI. That would be as silly as creating road rules by listening to motor engineers but not drivers. And inviting corporations into these discussions would be like having car salespeople in the room and still no actual drivers or, for that matter, the pedestrians who get mown down when the brakes fail. Yet, in viewing the AI challenge as a mechanical question of intellectual property, the federal government is in danger of doing just that as it considers the ownership of artistic works fed into and spat out of the AI meat-mincer.
It would be regrettable if scientists and corporations are given the greatest sway in making laws about art. As regrettable as artists making laws about science or business, which of course would never happen in Australia. Composing music or writing or painting or sculpting are real labour, but it is not an efficiency problem waiting for a mechanised solution. Many authors still work with pen and ink; it is inefficiency, raw manual labour, that often produces artworks of genius.
When I was thinking AI could relieve some of the grunt work from the writing professions, I was being short-sighted. As the writers on strike in Hollywood are stubbornly pointing out, AI will first replace junior careers. Gruntwork is how they learn their trade. You let AI do their job, you are eliminating the apprentice. For a futuristic tool, that’s a pretty good way to eliminate the future. It is an(other) assault on the young. Every big stride, in every walk of life, builds its foundation upon apprenticeships.
Nor is lawmaking a form of labour that can be outsourced to machinery. The typing part can be improved by predictive text, but not the thinking. Lawmaking is artistry too, requiring touch and instinct. I hope the lawmakers remember this when they are considering the future of other forms of art and work.
Malcolm Knox is a journalist, author and columnist for The Sydney Morning Herald.