The cells are cultivated using donated stem cells, which grow on a silicon chip with 59 electrodes. These electrodes send electrical signals to the cells and, in turn, «read» their responses. The entire system is kept at a stable temperature of 37ºC (the same temperature as the human body).

This system uses a reward-and-punishment software to train the neurons. This same principle is used both in our brain and in AI. For example, in our brain, when we suddenly eat ice cream, we feel «pain,» warning us that the action is harmful to our body. In computing, the system is rewarded when it performs positive actions and punished when it does negative ones. Through trial, error, and repetition, the AI learns, for instance, to play chess.

So, what’s the biggest advantage of using biological computers over conventional ones? The short answer is energy efficiency. The brain functions with just 20 watts of power, meaning that, for example, a rack of 30 CL1 computers needs the same energy as a toaster, while any other AI system uses millions of watts for training.

One example of a use case that seems more practical to explore with CL1 is testing medications that affect neurons. With the interaction between neurons and chips, results could be measured directly instead of attempting to simulate reactions.

This technology will begin to be commercially available in July for $35,000 USD.



Sources:
https://corticallabs.com/
https://www.youtube.com/watch?v=GQ5s0ID4vbY
https://www.youtube.com/watch?v=6nYpeS1fp6k