Scientists propose to develop a biological computer powered by millions of human brain cells that they say could outperform silicon-based machines while consuming far less energy.
The international team, led by Johns Hopkins University in Baltimore, published in the journal Frontiers in Science on Tuesday a detailed road map to what they call “organoid intelligence”. The hardware will include arrays of brain organoids — tiny three-dimensional neural structures grown from human stem cells — connected to sensors and output devices and trained by machine learning, big data and other techniques.
The aim is to develop an ultra-efficient system that can solve problems beyond the reach of conventional digital computers, while aiding development in neuroscience and other areas of medical research. The project’s ambition mirrors work on the more advanced quantum computing but raises ethical questions around the “consciousness” of brain organoid assemblies.
“I expect an intelligent dynamic system based on synthetic biology, but not constrained by the many functions the brain has to serve in an organism,” said Professor Thomas Hartung of Johns Hopkins, who has gathered a community of 40 scientists to develop the technology.
They have signed a “Baltimore declaration” calling for more research “to explore the potential of organoid cell cultures to advance our understanding of the brain and unleash new forms of biocomputing while recognising and addressing the associated ethical implications”.
Developing organoid intelligence into a commercial technology could take decades, Hartung conceded. On top of the scientific challenges there are ethical concerns about creating “intelligence in a dish” that can learn, remember and interact with its environment — and could develop consciousness even in rudimentary form.
An “embedded ethics” approach had been in place from the project’s launch, said Hartung, adding: “All ethical issues will be continuously assessed by teams made up of scientists, ethicists and the public.”
Madeline Lancaster, a brain organoid researcher at the Laboratory of Molecular Biology in Cambridge, who is not connected with the project, was sceptical about its ambitions. “This is really very much science fiction and, while intriguing, the science just isn’t there yet,” she said. “There are huge hurdles to overcome in order to do what the authors propose.”
Karl Friston, neuroscience professor at University College London, who is not involved in organoid intelligence, was more positive. “It is definitely an idea worth pursuing,” he said. “There will be many baby steps ahead but the direction of travel could be revolutionary.”
One necessary step, Hartung said, was to enable individual organoids to grow larger by finding a better way to suffuse them with nutrients in laboratory dishes. These tiny neural constructs need to be scaled up from about 50,000 cells today to around 10mn to help achieve what scientists would recognise as organoid intelligence.
Researchers are also developing technologies to link organoids together and communicate with them, sending them information and decoding their “thoughts”. Hartung’s lab has tested an interface, “a flexible shell that is densely covered with tiny electrodes that can both pick up signals from the organoid and transmit signals to it”.
One reason for turning to biological computing is that the brain processes and stores information so efficiently. The world’s most powerful supercomputer, the Frontier machine at Oak Ridge National Laboratory in the US, which became operational last year, matches a single human brain for processing power — one exaflop or a billion billion operations per second — but consumes a million times more energy.
The first applications of organoid intelligence will be in neuroscience and medicine. Scientists are already making brain organoids from stem cells derived from patients with neurological conditions, to compare with healthy individuals and assess their response to drugs. Organoid intelligence would boost research into the cognitive impairment caused by brain diseases — and its prevention.
While the technology may take decades to deliver biocomputers powerful enough to compete with conventional silicon or quantum systems in the provision of functionality such as artificial intelligence, proponents of organoid intelligence point to its immense and unpredictable potential.
“I hope that we see things which are not just a copy of normal brain development,” said Hartung.