ҹ1000

Combatting deepfakes is an evolutionary arms race

Disinformation is far older than humans. Lessons from evolutionary biology can help defend against it today, says Jonathan R. Goodman

THE recent “Willy’s Chocolate Experience” in Glasgow, UK, brought children to tears, angered parents and was a source of mirth for those who read about it. After seeing online advertising for an apparently lavish event, people paid up to £35 each to attend. But it wasn’t as promised: instead of wondrous creations and an abundance of chocolate, families arrived at a mostly abandoned warehouse featuring a sad-looking bouncy castle and confused actors.

When news networks into the event, it emerged that the marketing materials – and the actors’ scripts – were produced by artificial intelligence. The advertising created the impression that the experience would be worth the money, leading to a difficult question: in the era of AI, how can we tell whether we are being tricked?

Misleading ads are just one example of a veracity problem now endemic in our digitalised world. Yet even as deception grows more sophisticated, countermeasures are being developed: the BBC is a tool from Intel known as “Fakecatcher” that analyses changes in facial blood flow to detect, for example, AI-generated ads that may feature a famous .

There is an arms race going on in the digital world. But the competition between deception and honesty is as old as life on Earth. Understanding that evolutionary conflict can help us beat those who aim to lie to us.

Disinformation certainly predates the existence of humans. Cancer cells, for example, into treating them as a normal part of the body, rendering white blood cells useless against them. The body can’t eliminate rogues if it can’t tell friend from foe.

We see disinformation strategies among animals, too. lay their eggs in the nests of other birds. The foster parents are then deceived into raising the young of others at the expense of their own.

More complex are cases where primates seem to intentionally communicate false information to deceive. One study that wild tufted capuchin monkeys may shriek as though a predator is nearby to trick their fellow primates into running away from food. They then take the food for themselves – using, like a cancer cell or a cuckoo, disinformation to further their goals.

This links to what is known in biology as the . Derived in spirit from Alice in Wonderland, this effect refers to a case where two organisms compete to outmanoeuver each other. Bodies compete with cancers. Cuckoos compete with host birds. Primates compete with each other. In each clash, a new strategy by one party forces the other to adapt – or face elimination in the cold game of natural selection.

Disinformation today is the cultural equivalent, a more complex version of the false monkey shrieks. The only difference is the level of sophistication and the number of people involved.

The good news is that we know enough about evolutionary thinking to predict which steps bad actors will take next. We do this already with cancer: we have novel ways to lure it into evolving a weakness that we then exploit, a process called ““. Similarly, we know someone will try to outwit the tools being developed to counter deepfakes. If Intel’s tech is based on blood flow, we can predict what the next form of deepfake will look like, then pre-empt it. And so on.

Evolutionary biology in part describes ancient games of dishonesty and betrayal. We need to use our understanding of arms races to make traps of our own – disinformation traps. Armed with knowledge, we can go to war.

Jonathan R. Goodman is a research associate at Cambridge Public ҹ1000, UK

Topics: AI