A team of Australian scientists collaborating across academia and private industry has just received a three-year grant for their work growing brain cell cultures capable of communicating with machines.
Over the past two years, the team has already succeeded in teaching a brain cell culture of about 800,000 neurons how to successfully play the 1970s video game Pong from its petri dish.
The $600,000 grant was awarded by the Australian Government’s Military and Intelligence Community and will be administered by the Australian Research Council.
Pong project co-investigator, theoretical neuroscientist Karl Freeston, said last October, ‘The beautiful and pioneering aspect of this work relies on equipping neurons with sensation: feedback.’
‘And crucially,’ added Professor Friston, given brain culture, ‘their ability to act in the world.’
An image courtesy of Cortical Labs shows the team’s ‘dishbrain’ under a microscope with different types of cells with fluorescent markers: green marks neurons and axons, purple marks neurons, red marks dendrites. Multiple markers appear as yellow or pink
About two years ago, the Australian team that won the $600,000 grant had already succeeded in teaching brain cell cultures of about 800,000 neurons how to successfully play the ’70s video game Pong. Dishbrain figured out the game faster than the AI in just under five minutes
‘Remarkably, cultures have learned how to act on this to make their world more predictable,’ Freeston said in a press statement.
‘It’s remarkable because you can’t teach this kind of self-organisation; Simply because – unlike a pet – this tiny brain has no sense of reward and punishment,’ he explained.
The new project advancing the technology will be led by psychologist Adil Raji of Monash University in Australia, where Raji also heads the school’s Computational Neuroscience Laboratory.
The research was carried out in partnership with Melbourne, Australia-based start-up Cortical Lab, as well as University College London, where Freeston is based.
Through the National Intelligence and Security Discovery Research Grants (NISDRG) program, Australia’s Office of National Intelligence and its Department of Defense National Security Science and Technology Center award up to $18 million in high technology grants each year.
Notably, the $600,000 AUD grant awarded to these researchers is at the high end of those awards, which range from $400,000 to $600,000 per grant.
Their proof-of-concept success in teaching brain cell cultures to play pong was first made public in December 2021, a breakthrough that proved faster than computer-based AI.
‘The amazing part is how quickly it learns,’ Brett Kagan, Cortical Labs’ chief scientific officer, told New Scientist that month, ‘in five minutes, in real time.’
The fastest a fully silicon-based, computer AI can manage is 90 minutes to get a handle on playing Pong.
‘It’s really an amazing thing that biology can do,’ Kagan said.
The system, called ‘dishbrain’, consists of brain cells grown on microelectrode arrays that can stimulate both cells.
Researchers at Monash University and Cortical Labs have named their culture of brain cortical neurons, grown on an array of computer-connected microelectrodes, ‘dishbrain.’
As Kagan puts it: ‘We think it’s fair to call them cyborg brains.’
From the virtual world where pong is being played, feeds from the electrodes help the mini-brain learn how to operate the virtual paddle.
During a game of pong, patterns of activity across neurons are determined by the mini-brain when the paddle moves left or right.
‘We often refer to them as living in the matrix,’ Kagan said. ‘When they’re in the game, they believe they’re the paddle.’
With their new research funding, to judge from Raji’s statement through Monash University, the group hopes to revolutionize not just computing but the entire digital economy through Dishbrain’s advances.
‘The results of such research provide Australia with a significant strategic advantage, having important implications in multiple areas such as, but not limited to, planning, robotics, advanced automation, brain-machine interfaces and drug discovery.’
Speaking as the project’s leader, Raji believed Australia’s national security sector was in favor of their brain cell-computer interface project because they saw its potential in many areas where traditional AI was failing or too slow to advance.
Self-driving vehicles, fully autonomous drones and more symbiotic wearable devices require ‘a new kind of machine intelligence capable of lifelong learning,’ Raji said.
‘In the future the capabilities of these new technologies may eventually exceed the performance of existing, purely silicon-based hardware,’ he said.