One of the most daunting problems that current AI systems face is that they are too energy-intensive. Scientists have been working diligently to find a solution to this conundrum. If only they could replicate how our brains work. Nature’s creation might solve their problem. That’s why scientists have been closely studying the brainpower of biological neurons that power the human organ.
And the new inventions are being modelled after this biological model. Chip giant Intel recently introduced neuromorphic chips which used the human brain as a model. In other major advances in neuroscience, scientists developed artificial neurons and were also able to successfully connect biological and artificial neurons online.
Taking all these endeavours one step further, now an Australian startup, Cortical Labs based in Melbourne, is building miniature disembodied brains using a combination of biological neurons and a specialised computer chip. The company founded last June has received about $610,000 in seed funding from a prominent Australian venture capital firm.
To begin with, Cortical Labs employed two methods to create this innovative hardware — the first one involved the use of mouse neurons extracted from embryos, while the second one embraced a technique in which human skin cells are converted into stem cells before transforming them into human neurons eventually.
“The company is planning to build technology that harnesses the power of synthetic biology and the full potential of the human brain in order to create a new class of AI that could solve society’s greatest challenges,” it said in an announcement.
The next step involved embedding these neurons in a nourishing liquid medium on top of a specialised metal-oxide chip which contained 22,000 tiny electrodes — enabling programmers to provide electrical inputs to the neurons and also sense their outputs.
The mini-brains created as a result have the processing power of a dragonfly brain — thus being able to play the old Atari arcade game Pong. This is a significant benchmark as the same game was used by the famous artificial neural network DeepMind to demonstrate the performance of its AI algorithms in 2013. Google acquired the company in 2014.
Karl Friston, a leading neuroscientist at University College London, known for his work on brain imaging, was given a demonstration of the Cortical Labs’ technology earlier this year and was impressed with their achievement. Certain aspects of Cortical Labs’ system are based on Friston’s research.
Friston expected that his work would be used to build more efficient neuromorphic computer chips, as we saw Intel do recently, but he didn’t expect the integration of biological neurons with today’s standard computer chips saying, “What this group has beenable to do is, to my mind, the right way forward to making these ideas work in practice.”
However, Cortical Labs is not the only company working on biological computing. A California-based startup Koniku, has developed a 64-neuron silicon chip, built using mouse neurons, that can sense certain chemicals. They expect that military and law enforcement can used their invention to detect explosives. Also, MIT researchers have taken an ingenious approach by utilising a strain of bacteria in a hybrid chip to build a computing system.
AlphaGo, DeepMind’s deep learning system, which defeated the world’s best human player in that ancient strategy game – Go – in 2016, used up one megawatt of power while playing the game, as compared to about minuscule 20 watts of power that a human brain uses to play the game. While the achievement was stupendous for the AI system, it was overshadowed by the enormous amount of power need to run the system.
If Cortical Lab can scale their recent innovation, it can provide a solution to one of the most long-standing problems plaguing the futuristic technology.
First published on Medium