THE washing-up pile wobbles precariously as you balance another saucepan at its summit. For a second, it looks like the whole stack will come down. But it doesn’t. Swiftly, instinctively, you save it.
Congratulations – not just on another domestic disaster averted, but also on showing a peculiarly human genius. Octopuses rival our dexterity, New Caledonian crows have a frighteningly clever way with tools and chimps beat us in tests of short-term memory. But no other species can perform complex, real-time calculations of their physical environment and generate specific, actionable predictions quite like the ones that rescued your crockery. “It’s kind of amazing,” says artificial intelligence researcher from Google DeepMind in London. “To me it defies my ability to understand.”
In 2013, Battaglia and two colleagues showed that our inbuilt “physics engine” , software used in video games to generate a realistic playing environment. It is programmed with rules about objects’ physical behaviour, and uses limited real-time inputs (from a player in a game, from our senses in reality) plus probabilistic inference to generate a picture of what comes next. “What you have in your head is some means for running a simulation,” says Battaglia. “You make a 3D model of what’s around you and press the run button, it tells you what will happen. It’s a way to predict the future.”
Advertisement
Earlier this year, at Johns Hopkins University in Baltimore, Maryland, and his colleagues a task involving physics intuition – predicting how a tower of stacked wooden blocks would fall – and showed that the physics engine sits in specific brain regions. Areas of the motor cortex associated with the initiation of bodily movement consistently lit up during the first task, but not on a second, purely mathematical task, estimating the number of different coloured blocks in the tower.
“What you have in your head is a method to predict the future”
That was surprising at first, says Fischer. “But on the other hand it makes perfect sense: you don’t execute any action without mental models.” So our inbuilt genius won’t necessarily help us with physics as an academic discipline, which relies on different brain circuits. That much is clear in experiments where researchers get people to draw the predicted path of a falling object, says Fischer: their intuitions are completely off. But have them catch the same falling object, forcing them to engage their motor system, and they’re spot on.
There’s still a lot to learn about how we generate our simulations – not least given that our device’s power consumption, at around 20 watts, is less than a tenth that of a medium-range graphics card. “The type of processing we use is clearly vastly more efficient,” says Battaglia.
But we should be aware of our limitations, too. Our physics engine is programmed with the equations of classical mechanics, which describe the visible world around us – things like falling plates. It does not work so well on less obvious layers of reality. “Understanding electromagnetism and quantum mechanics, our instincts are not going to be so useful,” says Fischer. There, things don’t stack up so easily.
Read more: You are amazing – The 10 incredible superpowers in every human
This article appeared in print under the headline “You are… a physics genius”
