MELBOURNE: Australian researchers have trained lab-grown brain cells on a silicon computer chip to play the 1990s shooter “Doom” and say they are only scratching the surface of what the neurons might be capable of.
It is the science fiction work of biotechnology experts at Cortical Labs, who researched and developed the technology that harnesses the functioning of the brain’s networking system.
Each of the so-called “biological computers” contains around 200,000 live human brain cells, grown from stem cells taken from blood donations.
Having mastered the simple computer game “Pong,” in which you move a paddle up and down to send a ball across a screen, the brain cells have moved on to bigger things.
Initially, the neurons were at the “level of a beginner who had never played a video game before,” said Alon Loeffler, senior applications scientist at Cortical Labs. AFP.
“Doom” involves a chaotic 3D game world where the user must explore their surroundings and eliminate enemies, which is not an easy task for a group of cells.
“They were walking into walls a lot, shooting at them, rolling around and doing funny things like that,” Loeffler said.
“And eventually they started targeting enemies more regularly and correctly.”
It’s not the cleanest execution, though. A demon requires several attempts to slaughter, with shots in multiple directions before reaching the target.
But mind-blowing research shows that neurons can adapt to stimuli in real time and complete goal-directed learning, says Cortical Labs.
‘Scratching the surface’
The researchers converted “Doom’s” digital environment into patterns of electrical signals that the chip’s neurons could understand.
When an enemy appears, specific electrodes stimulate the neurons in the special chip called CL1, causing them to react.
Different patterns of neural activity produce specific responses, such as firing the gun or moving left or right.
Researchers monitor the electrical activity of neurons from a computer screen connected to CL1, represented by thousands of small dots.
From this data, the team adjusts its inputs to influence and train the neuron’s activity.
The CL1 is not limited to computer games: the chip can be coded to perform a variety of applications, from drug detection to AI-like machine learning.
“We’re just scratching the surface of what these neuronal cultures can achieve when integrated into systems like our CL1,” said Chief Scientific and Operating Officer Brett Kagan.
“Our neural cultures have been explored for a variety of tasks,” he said, from “robotics, AI-like real-time learning tasks, as well as healthcare, medicine, disease modeling, drug screening, and even personalized medicine.”
It’s not a “mad science”
Kagan describes the CL1 chip as “a more sustainable and powerful form of intelligence.”
The human brain runs on about 20 watts of power, a level of efficiency that silicon computing and artificial intelligence have not yet been able to replicate.
While it is “not intended to replace what AI does,” its goal is to “give us abilities we’ve never had before,” Kagan said.
The cells have a shelf life of six months and are not yet capable of producing consistent, programmable results.
But analysts say the project’s value could lie in its more sustainable energy consumption compared to regular chips.
“We need better ways to manage that energy and get higher levels of efficiency,” said William Keating, chief executive of semiconductor research company Ingenuity.
“This isn’t crazy science or a bunch of scammers. It’s real science and it’s making real progress.”
