AUSTRALIA’S Peter Miller, aka Scribbletronics. has been working with digital art for two decades. It’s a long way from his early days, tinkering with mathematical rules to generate procedural images, but Peter’s approach has evolved for the age of Midjourney, DALL-E, et al. Prompt-crafting today involves directing software to dip into its training database, using tags to identify and evoke concepts – what, say, does “worried” look like if we ask for a depiction of a worried man?

This article is featured on the AI Hub

"Sometimes fate is only a failure of imagination. The trajectory of automation is not only uncertain, it is unwritten."

“A lot of people forget that AI tools don’t have a spatial awareness like that of humans,” Peter explains. Instead of representing imaginary objects, the AI is exploring concepts.

“Ask for a dinosaur, it’s pretty straightforward – it looks for everything that’s tagged as a dinosaur and tries to make some kind of confluence of those things. If you ask it for Italian marble, it goes and looks at lots of images tagged as depicting Italian marble. The real power comes when you give it an instruction to make a dinosaur from Italian marble, and it looks at the intersection of the tagged materials, and does its very best to make a dinosaur that has the characteristics of marble. If you ask it for a worried dinosaur made of Italian marble in a supermarket, it may struggle because the intersecting concepts are that little bit more complex – but it’ll do pretty well.”

Peter acknowledges that AI training sets have appropriated work by other artists, representing something of a “poisoned chalice”, but connects it to wider histories around intellectual property and the sampling of music: “The question of appropriation was as significant then as it is for AI now, but most people were unaware of it because it affected only musicians.”

Where does that leave us ethically? Sci-fi author and social scientist Malka Older argues that “AI can serve as a cover for the choices of human authorities, slotting into a “role previously held by other amorphous concepts like ‘bureaucracy’” – a way of depersonalising decisions and avoiding accountability.

The late Joanna Russ, another science fiction writer, argued technology itself had become a guise for something else since the age of mass production:

“It is the entire social system that surrounds us; hence the sense of being at the mercy of an all-encompassing autonomous process that we cannot control. If you add the monster’s location in time (during and after the Industrial Revolution), I think you can see what is being discussed when most people say ‘technology.’ They are politically mystifying a much bigger monster – capitalism in its advanced, industrial phase.”

Carissa Véliz, a philosopher at Oxford’s Institute for Ethics in AI, argues against putting all the blame on capitalism, arguing that we should consider data and privacy “in terms of power, rather than money. Power isn’t just about wealth.”

She explains: “That connection between data and power is often obfuscated by those who benefit from it. Bertrand Russell had this insight that power behaves like energy, transforming from one kind to another. Enough economic power buys you political power; enough political power gets you military power; and so on. In the digital age, we have a new manifestation of power.”

Professor Véliz notes that we could have had a different kind of AI.

“Though machine learning is very impressive and has caused some truly exceptional advancements, it has limits and disadvantages. It’s not very good at identifying causation rather than correlation, for example. It can also leave us stuck in the past, as it typically uses historical data, perpetuating tendencies and biases we might prefer to get rid of. If we had more symbolic AI, or hybrid AI, we would need less data.”

The world could always be otherwise than it is. Sometimes fate is only a failure of imagination. The trajectory of automation is not only uncertain, it is unwritten. The fundamental question is less about technical capacity than ethics, authority, and accountability: Who will create the future – and with what ends in mind?