Source: Radio New Zealand
SPENCER PLATT
Researchers at Melbourne start-up Cortical Labs have taught their “biological computer” made from living human brain cells to play Doom.
They say it brings biological computers a step closer to real-world uses, such as drug-testing or robotics applications.
Cortical Labs synthetic biological intelligence scientist Dr Alon Loeffler told Midday Report it was the “first code-deployable biological computer”.
“We like to call it neurocomputer, made out of about 200,000 to 800,000 cells that were taken from stem cells and turned into brain cells,” he said.
“Then we had an early access user, a customer of ours, in one-week programme the game Doom, or a free version of Doom, without the copyright restrictions, so that the cells can navigate this environment and try and beat the game.”
He said the cells were very similar to what would be in a real-life brain.
Loeffler said while they were human brain cells, they were not taken from people’s brains, but rather from blood donations.
“We take blood donations from willing volunteers and donors and then our amazing biology team does some biology magic, which is science, but I think of it as magic.
“They turn these blood cells into stem cells, similar to what in the past you’d have to take out of embryos, but now you can just get them from skin cells or blood cells.
“Then those are converted to brain cells or cortical cells, which are then placed on a Petri dish, and we can record the electrical activity from the cells because they communicate via electrical signals, similar to how they would in the brain.”
In that sense, they were alive, he said.
‘Learning to improve over time’
Loeffler said because the system didn’t have sensory inputs such as eyes or ears, the question was how they would encode the information.
A lot of research had gone into that, he said.
“We’re still in the very early stages of understanding that, but the idea is, for example, in the Doom game, if there’s an enemy or demon that appears on the left side, you can send in an electrical input on the left side of the chip, and if it’s on the right side, you could send in an electrical signal on the right side of the chip.
“This is obviously a much more condensed version and simplified version, but then the response of the culture would then kind of tell the game or tell the controller what to do, to move to the left or to move to the right, for example.”
Loeffler admitted the computer was not very good at the game, but would outperform a model that shot randomly.
He said it was “learning to improve over time”.
Loeffler said there were several real-world applications it could be applied to, such as drug development and testing.
“You can test all sorts of different drugs on these cells, and they’ll perform much more similar to biological systems,” he said.
“They’re also much more similar to brains than animal models, so you can kind of remove the need for mice and chimpanzees and sheep in animal models. You could also potentially use them for robotics applications.
“It’s one thing that biological systems are really good at doing, which AI is terrible at doing, is navigating new and changing environments.”
He said if they could improve its ability to understand inputs, they would be able to navigate an environment in a more biological way.
Sign up for Ngā Pitopito Kōrero, a daily newsletter curated by our editors and delivered straight to your inbox every weekday.
– Published by EveningReport.nz and AsiaPacificReport.nz, see: MIL OSI in partnership with Radio New Zealand
