Cortical Labs has officially launched the world’s first “biological computer.” This computer integrates live human brain cells with electronic chips to create an inherently intelligent biological computing system. With this achievement, the company shifts artificial intelligence from relying on silicon chips to power artificial neural networks, to utilizing natural neural networks to perform intelligent tasks.

The Australian company named the biological computer CL1 and officially announced it at MWC Barcelona, describing it as a “body in a box” with the potential to revolutionize artificial intelligence and robotics. The company asserts that this new type of computational intelligence, using brain cells, is more dynamic, sustainable, and energy-efficient than any existing AI. Despite the complexity of caring for neural cells in a laboratory setting, they hold promising and worthy-of-exploration capabilities.

The company will begin shipping units in the second half of this year, with each unit costing approximately $35,000. Users will also be able to access and use it remotely via the cloud to train intelligent systems using lab-grown brain cells.

The computer consists of a network of microelectrodes onto which neural brain cells are cultured. The entire system is housed within a unit equipped with pumps and temperature controls to maintain cell viability.

How is it trained to perform intelligent tasks?

Researchers connect the electrodes, which interact with the cells, to a computer program that simulates the desired task. The program sends inputs to the cells as electrical pulses and receives outputs from them as electrical pulses via the electrodes beneath them. When the cells successfully perform the required action, they receive positive reinforcement through the transmission of electrical pulses that strengthen this behavior, thus reinforcing the connections between the cells that contributed to this success. If the cells fail to perform the required action, they receive negative reinforcement through a different pattern of electrical pulses, prompting them to modify their behavior.

Over time, the neural cells adapt based on the feedback they receive, and they can improve their performance by strengthening the neural connections associated with correct actions and weakening those associated with incorrect actions. This process allows the neural cells to learn and achieve goal-oriented behavior.

In 2022, the team published research on training approximately 800,000 human and rodent cells to play the game Pong. These cells successfully learned to move the paddle to direct the ball appropriately, paving the way for the creation of this biological computer. This computer serves as a foundation for a new phase of innovation, where each user contributes to the system’s development with new ideas, potentially opening up vast possibilities for enhancing its capabilities.

Company researchers found that neural cells train faster than artificial neural networks and consume less energy and data to master tasks. Based on these findings, they anticipate that the biological computer will one day be able to produce intelligence comparable to large language models like ChatGPT, but at a lower cost.

Furthermore, these cultured cells can be used as a model to study the effects of drugs and neurological treatments, accelerating the drug development process and contributing to a deeper understanding of the brain and its mechanisms and development.