A small number of companies are working on biological computers
Floriana/Getty Images
Data centers use huge amounts of energy and chips are in demand – could brain cells be the answer? Australia-based startup Cortical Labs has announced it is building two “biological” data centers in Melbourne and Singapore, stacked with the same neuron-filled chips it has demonstrated can play Pong or Doom.
Cortical Labs is one of the few companies developing biological computers, consisting of neuronal cells connected to microelectrode arrays that can stimulate and measure the response of the cells when fed data. Earlier this month, the firm demonstrated that its flagship computer, the CL1, could learn to play the game Doom in a week.
Now, Cortical Labs has revealed two data centers that it plans to build. The first, in Melbourne, will contain around 120 CL1 units. The other, being built in partnership with the National University of Singapore, will initially house 20 CL1s, but the company hopes it will eventually house 1,000 units in a larger data center, subject to regulatory approval. Cortical Labs says this will allow it to expand its cloud-based brain computing service.
Biological computers like CL1 are being built and tested by research groups around the world, but they are often difficult to build and not easy for others to use, says Michael Barros of the University of Essex, UK. “We spend a lot of money and sweat building these (systems).”
“What (Cortical Labs) is doing is essentially making its biocomputer available at scale,” says Barros, who already uses Cortical Labs cloud services as part of his research. “They want to be the first to do it.”
Although these systems can be trained for relatively simple tasks, such as gaming Doomthe exact way these neurons work and how best to train them to perform tasks such as machine learning is still unclear, says Reinhold Scherer, also at the University of Essex. “Having access to this allows you to explore learning, training and programming,” he says. “You don’t program neurons like regular computers.”
Cortical Labs claims that its data centers will also require far less energy than typical computing systems, claiming that each CL1 needs around 30 watts, rather than the thousands of watts that a state-of-the-art conventional AI chip requires.
“When we scale up and have these as whole rooms, just like you have now with data servers, there can be huge power savings,” says Paul Roach of Loughborough University, UK. There are other resources that biological data centers may need, such as nutrients to feed and keep neuronal chips alive, but it should require far less cooling than conventional computing, he says. “The amount of energy saved with (Cortical Lab’s) numbers is quite conservative.”
However, the technology is still at an early stage, says Tjeerd olde Scheper of Oxford Brookes University, UK, who has worked with a competing biological computing company, FinalSpark. “Is it going to work like people might think? No, we’re still at the beginning of this development.”
It’s hard to make a direct size comparison, since CL1 chips can’t do conventional calculations like a regular silicon-based AI chip can, but the proposed biological data center will have hundreds of biological chips, compared to the hundreds of thousands of graphics processing units (GPUs) seen in the largest AI data centers.
“I think it’s a very long way from production-ready. It’s a very big step from a small network playing a computer game to an LLM,” says Steve Furber of the University of Manchester, UK.
One of the remaining problems is that it remains unclear how to store the results of training these neurons in some form of memory, or how to run actual computational algorithms on them, rather than training them for specific applications such as video games.
Another challenge is how to retrain the neurons once they have completed a particular task. “What they’re trained in is lost when the culture ends, so there has to be a real retraining,” says Scherer. “Then it is not an optimal solution to keep a technology running if you need retraining every 30 days.”
Topics:
