James Bridle: rethinking artificial intelligence on natural lines
Artist and writer James Bridle argues that by sticking to human models of intelligence, we're limiting AI's potential.
James Bridle is a writer, artist, and technologist. In the second episode of the NEXT – SHOW series 4, they explored how our models of AI might be deeply flawed, and how a new approach could solve some of humanity’s biggest challenges.
Watch the complete episode
James has a new book out, called Ways of Being, about “AI, non-human intelligence, ecology, biological computing, more-than-human relations, and much else”. It started out from the same place as their previous one — New Dark Age — in that it looks at the way digital technology has developed and how that has not really worked out for us as we thought it would. And that causes problems. They wanted to expand on some of those ideas and, in particular, the emergence of artificial intelligence. AI, and other technologies which make very broad claims about what they can do for the world, are interesting because they can have impacts which can be surprising, risky and perhaps even dangerous.
“I found myself questioning what we were doing wrong,” Bridle said. “I focus on intelligence, and why we think that artificial intelligence might look like the things we’re building – and could we do things in a different way.”
Bridle realised that we have a very narrow idea of what intelligence could be, derived from our own intelligence. If you look at how intelligence is done by other beings we share the planet with — animals, plants, microorganisms of all kinds, and even whole ecosystems — and took them seriously, we could learn from them. We could sketch out a different path for both technology and the entire planet.
Redefining intelligence in a non-human way
Bridle ended up broadening out the common definition of intelligence as they explored these ideas in the book. For example, intelligence can be a kind of embedded knowledge, which we only have just come to understand.
There’s a type of plant called hyperaccumulators, which draw metals up from the soil into their leaves and stalks – and we can harvest them for metals, like a crop. It’s metals extraction without all the damage and pollution of mining.
The scientists Bridle’s been working with have found plants like this all over the world, which have learned to do this by evolving in a very particular place over millennia. They have an embedded knowledge of the place that humans are only just beginning to recognise. These plants are using this knowledge to survive — and we’re beginning to work out how we could work with them, for a more sustainable way of obtaining the things we need.
The embodied wisdom of a slime mould
Then look at slime moulds. Yes, slime moulds. They’re an extraordinary kind of microorganism that sits somewhere between plants and fungi. But they’re very good at mapping out networks and figuring out the most efficient routes. When researchers put them to the test with the travelling salesman problem, they could solve it better than any super-computer we have yet developed.
“For them, it’s just a party trick,” Bridle said. “We’re interested in logistics and transport networks, but they’re probably interested in all kinds of other things, and if we work out what other questions to ask them, we could learn all sorts of things.”
If you look at how we’re designing AI now, they’re insular and competitive. This is because they’re designed to be single, fixed machines, working through predetermined routines. And they’re designed on adversarial processes. If you think about that in evolutionary terms, you’re building a highly competitive organism – but competition isn’t the only model of intelligence.
Some creatures have intelligence spread throughout their bodies, like octopuses and slime moulds, and sometime intelligence exists in the relationship between bodies. So, we could look at intelligence as distributed, and based on co-operation and relationships, instead. AI might need to become more like non-human natural systems, and explore models of cognition from plants and animals.
Artificial intelligence isn’t artificial
The Internet of Animals project is using large-scale tracking of animals to make visible aspects of the natural world that we haven’t seen before. It can be used to predict earthquakes, by spotting animals making unusual movements just before the quake. Those patterns aren’t discernible by humans – but machine learning systems trained for stock market trading could pick it up. This is a great example of a more collaborative model of AI.
AI can be a gateway to understanding the world differently. We’ve tended to view human intelligence as the pinnacle of intelligence, and been reluctant to admit that other forms exist. But we’re slowly becoming aware of all the other forms of intelligence around us. The drive towards AI could well end up helping us understand these multiple forms of intelligence better.
In fact, Bridle hesitates to describe AI as “artificial”. It’s as much a product of this Earth and evolution as we are, in their eyes. Nothing we make is purely man-made. All the computers we use are made of minerals and electrons that are and always have been part of the world.
Signs of life — and intelligence
Bridle’s current exhibition — Signs of Life — is quite different from previous works. They’re very solid, working objects, like a windmill, solar panels, and a huge battery made of lemons… They all work, and that’s intentional. Bridle feels that art about ecology should reflect its functional nature. There’s a strong DIY ethos behind the art. If you understand how something works, you can build it — and that gives you more power and agency. The more people who understand, the more power is distributed. And that resists the tendency of tech to centralise power, Bridle argues.
For example, the solar panels have patterns of small sea creatures engraved on them. Why? Because research has shown some patterns engraved on solar panels can improve their efficiency, by trapping certain frequencies of light that would otherwise be reflected. The art moves our conception of the objects beyond the merely functional.
- Human intelligence may not be the best model for AI
- Natural systems often have an embedded intelligence that can solve problems we find challenging
- Finding new ways of developing AI on non-human models of intelligence may unlock AI’s potential
- AI is less artificial than we assume…
This is a summary of an interview with James Bridle conducted by Ina Feistritzer and broadcast on the NEXT Show on 16th June 2022. You can catch up with James and their work on Twitter and their website.