How AI could turn us all into Dr Dolittles
Throughout the 1960s and 1970s, thousands of whales were being killed every year, mainly for pet food, soap and oil. As a result, these majestic creatures were heading for extinction, fast. Humpbacks, for example, had numbered around 100,000 in 1900 but by 1970 there were fewer than 7,000 left alive.
Remarkably, all of a sudden the world woke up to this tragedy and in 1982 a global ban on whaling was agreed by more than 60 countries. This radical shift in public opinion can be traced to a little-known military engineer named Frank Watlington, stationed in a secret US military base on the Bermuda coast. Watlington was using deep-sea microphones to listen out for Soviet submarines, and in the 1950s he became the first person in history to document the mesmerically haunting sounds of whales singing underwater.
For a long time, the tapes were kept secret by the US military but they were eventually released in the early 1970s as an album called Songs of the Humpback Whale. Now that the bellowing of whales can be heard in every cheesy health spa and acupuncturist’s waiting room, it’s hard to describe the impact that this album had around the world. Hearing these calls for the first time encouraged millions of people to think afresh about these creatures and helped inspire the whaling ban just a decade later.
As we look ahead to 2026, there are many reasons to feel dismayed about the state of the world but one cause for optimism is that advances in artificial intelligence mean we might be on the brink of another pivotal moment in our thinking about the natural world, much like the one that Watlington helped usher in.
Researchers have found that marmosets have names for each other
This festive season, while most of us are lounging on sofas and scoffing chocolate oranges, groups of scientists are scattered around the globe — suspended in bat caves, diving to far-flung coral reefs and camping out in damp rainforests — busily documenting the myriad ways that other species communicate with one another.
The reason these researchers are out in the field is that there is a frantic race under way to collect enough audio and visual recordings to train AI models to decode what these creatures are saying. It’s an exercise broadly analogous to the way in which technologists used trillions of written words pulled from the internet and elsewhere to build ChatGPT and other AI systems.
Jeremy Coller, a British financier spending millions to promote animal welfare, is helping to turbocharge this scientific work by offering $10 million to the first academic group that pulls off “fluent two-way animal communication”. In other words, using machine learning not only to figure out what these creatures are expressing but also to enable us to talk back to them, in their language. The hope, of course, is that if we’re able to talk back and forth, it’ll bring us closer to them and perhaps even help us appreciate our kinship with the natural world.
According to experts working in the field, achieving this ambition might be five years away or more. But to help spur progress towards this objective, Coller — through the wonderfully named Dolittle Prize — is also handing out $100,000 to the research group that has made the biggest contribution to deciphering animal language over the past 12 months.
If you look at this year’s Dolittle shortlist, you’ll find some wonderful examples of how AI and scientific curiosity are deepening our knowledge about the diversity of life on Earth. Researchers using AI to study marmosets — tiny monkeys no larger than your hand — have discovered for the first time that these miniature creatures have unique names for each other, so they can call out to others on an individual basis.
Another research group looked at cuttlefish and filmed them waving their four front arms around, which when analysed by AI software turns out to be a form of sign language. (Although scientists haven’t yet cracked the code of what these cephalopods are saying to each other.)
AI is already being used with animals in more dubious ways
This year’s Dolittle winners were a group of biologists who have spent years taping the clicks and calls of dolphins and, thanks to AI, they have been able to identify “signature whistles”, which are essentially the mammals vocalising their own names, as well as “query calls”, used to point others in the pod towards something of interest.
All of these projects are studying wildly different species but what they have in common is the attempt to use technology to change the way we relate to animals, in a positive way.
This agenda is worth cherishing — in part because AI is already being used with animals in more dubious ways. Take AI dog collars for example, which are already on the market in the US. They can monitor a dog’s behaviour and administer electric shocks if the pooch is acting “badly”. I don’t know about you but the idea of humans outsourcing the administration of pain to an autonomous machine feels like a moral rubicon that ought not to be crossed.
So it’s a blessing that the London School of Economics has established a Centre for Animal Sentience (also funded by Coller) that brings technologists together with philosophers to push for the ethical use of AI in the natural world.
When I spoke recently to Professor Jonathan Birch, who runs the new LSE centre, I asked him what he hoped might be achieved if AI ends up decoding the mysteries of animal language. He thought for a while, then said: “With greater understanding comes greater compassion.”
I can’t think of a better wish for 2026 than that.
No comments:
Post a Comment