Why understanding human brain efficiency could help solve the artificial intelligence energy crisis
Examine, a free weekly newsletter that covers science with a skeptical, evidence-based eye, is delivered every Tuesday. You are reading an excerpt – Sign up to get the full newsletter in your inbox.
Agreeing to participate in a brain experiment seemed fun. Until he sees five scary syringes filled with blue goo.
I was told it would only hurt a little.
The experiment I was about to conduct revealed something fundamental about our brain function: how we respond to surprises.
Dealing with the unexpected has been a critical part of survival throughout our history: Cave lion! Avalanche! E-bike!
So how does our brain process the constant onslaught of sensory input while responding efficiently and sometimes without thinking to the risks and rewards of daily life? This question has inspired opposing thinking for decades.
Scientists who put an electrode cap on my head and injected droplets of conductive gel into my scalp from those scary syringes helped resolve the debate.
Its results take us into the minds of World Cup forwards and goalkeepers, and could one day even help reduce the energy demands of the monolithic data centers taking over our suburbs. Here’s how it went down.
Do our brains prioritize surprises?
The lead author of the new study, Dr. Scientists like Reuben Rideaux Journal of NeuroscienceWe are most amazed at how our brain works. Hyper-efficient machines around the world.
Our noodles can perform billions of billions of mathematical operations per second using the equivalent power of a small aquarium pump.
“The more we learn about the brain, the more we realize how amazing what it does is,” says Rideaux, of the University of Sydney. “What are the design principles that enable the brain to work so well and efficiently?”
Scientists who study the brain are divided into two opposing camps on the question of surprise.
One view is that the brain activates more strongly when responding to predictable events to ensure that we act appropriately to our most common daily triggers.
The other side suggests that brain power is actually suppressed to save energy during predictable events, and neurons fire more when something unexpected happens.
It makes sense to spend more energy on predictable events because these events occur more frequently. But during a surprise, the opportunity to gather new information is greater.
Does the brain prioritize energy for the expected or the unexpected?
“There’s actually been a lot of empirical evidence for both in the last decade,” says Rideaux.
The test begins
The skull is dotted with 64 flashing electrodes, the jaw is nestled into the head frame, my eyes focus on a black donut racing across the gray computer screen as the camera calibrates the movement of my pupils.
Then the test begins.
Inside a white circle, gray dots, wonderfully called Gaussian spots, flash in seemingly random dots. My job is to indicate on the keypad whether the dot appears to the left or right of the center of the circle.
The flashes are fast and the game moves quickly; I’m racing against the clock to be as fast and accurate as possible.
Every once in a while, I test to see if I can remember the exact location of the last gray drop that disappeared.
Throughout the 30-minute test, these seemingly random spots follow a hidden pattern. They often gather on one side of the circle, then jump to the other side.
The experimenters are secretly teaching my brain to expect the dots to be on one side and the other, and then testing whether I can remember their positions better when they appear where I expect them or in a surprise position.
My results, study co-author, syringe user, and cognitive neuroscientist Dr. Reviewed by Dominic Tran.
My brain waves, pupil dilation, and gaming results, along with data from 40 other study participants, show that we respond more quickly to predictable events.
So fast, in fact, that our brains actually start reacting to something we expect before it happens: an observation the team was able to make using cutting-edge neuroimaging techniques.
This explains why a tennis player begins to react to a 170 km/h serve before the ball even flies off the opponent’s racket, saving precious milliseconds by predicting where the serve will land based on playing experience and body language.
But the data also shows that my brain waves go up even more when I see a drop in a surprise location. I, along with the other participants, remembered the locations of these unexpected flashes more accurately.
The data suggest that the brain slows down to capture more information at unexpected moments, such as performing a “software update,” so it is better prepared for the future, Rideaux said. As a result, we remember these surprises much better.
So, does our brain prioritize the predictable or the unexpected? The answer is both.
The brain switches between a fast automatic mode and a slow data capture mode in the blink of an eye, 50 to 100 milliseconds.
Both pathways are effective for the brain: The brain chooses to simply react quickly with fewer details, or to respond more slowly but with vivid details and better memory of the event.
“The brain has its cake and eats it too,” Rideaux said.
World Cup brain science
Picture it this way: Sometimes you can barely remember doing it on the way home from work or school. Your brain reacts quickly to known turns and stop signs on the road with very few resources.
But if you are rear-ended by Mr. Whippy’s van on your way home, there will no doubt be memories lodged in your brain. You’ll be able to remember that moment in bright technicolor for years to come.
You can also imagine a World Cup goalkeeper during the penalty shoot-out. Based on his knowledge of the striker and the opponent’s movement, the goalkeeper begins to react to a kick before the shoe even meets the ball. If he jumps to the left corner where he waits for the ball and saves the goal, the incident will be blurry.
But if the striker fakes and lobs the ball into the middle of the net, our goalkeeper will likely remember every detail of that ill-fated attempt in chilling detail, right down to how it hit the grass and where the ball buried itself behind the goal.
Less voracious AI
Rideaux said that the findings revealing how the brain works so well could help solve the artificial intelligence energy crisis.
“AI has done incredible things. But it requires a huge amount of energy,” he said.
“We hope that by understanding the principles of biological computing that allow them to process information truly efficiently, we can apply them to artificial computing, neural networks, to improve their performance.
“This is really important for advancing AI and also reducing the rapidly increasing carbon footprint of data centers being built to meet the increasing energy demands of AI tools.”
The Examine newsletter explains and analyzes science with a rigorous focus on evidence. Sign up to receive it every week.
Angus Dalton is science correspondent for The Sydney Morning Herald.Connect with: X or email.
