Hybrid computer combines brain cells and computer chips

Eindhoven University of Technology announce Scientists want to build the first “Brain-on-Chip AI” computer. According to Regina Lutge, one of the team leaders on the project, the big advantage is that such a system uses much less power for computations than “normal” computers.

Computers can be used for machine learning, a form of artificial intelligence (AI) where systems can learn based on processed data. So says Bert de Vries, who is also the captain of the team machine learning It is currently still very expensive, both in terms of money and energy consumption. “Hybrid computing using brain cells could be the answer.”

Thomas Bayes

To achieve this, the BayesBrain Project was created, with Bayes coming from 18th century English statistician Thomas Bayes. Today, the work of a mathematician is used, among other things, in the field of machine learning. The brain part of the project speaks for itself: it simply comes from the human brain.

Brain cells that talk to computer chips

To build Bayesbrain, scientists will first use about a thousand human brain cells to form neural circuits through the use of so-called Microfluidic system† Cells will be kept alive by providing them with water and other nutrients. After about three weeks they should be ready to plug into your computer chips.

Next, the scientists will use the computer part to solve a simple task, the inverse pendulum problem. This is a simple task that requires the computer to learn how to balance a stick on a moving platform. He. She Inverse pendulum problem It is already being used regularly in machine learning to teach a machine the desired and undesirable behaviour.

Once the silicon side of the BayesBrain has mastered this task, the researchers will gradually transfer the computing power to the side of the brain. Ultimately, the goal is for both sides to work together to efficiently solve the problem.

“The most energy efficient computer”

Technology is ultimately intended to be used to solve real-world problems. An example would be managing air traffic control at a large airport.

According to De Vries, if BayesBrain is working properly, this could be a huge step forward in the computer world. “Most importantly, this is a new beginning with the potential for revolutionary change. We plan to use the most intelligent system ever, the brain, as a component of what could be the world’s most energy efficient computing device.”