For years, computer scientists have been trying to use computer chips to simulate the human brain. We took a different approach: to build a new breed of computer chip by using real brain cells (neurons).

By combining traditional silicon computing with neurons taken from mice embryos or grown from induced human stem cells, we are the first in the world to have created a hybrid computer chip that is both alive and thinking. These chips are kept alive by being immersed in a nutrient-rich media that resembles the Cerebral Spinal Fluid (CSF) that bathes our brains.

While the original goal was to prove that neurons grown outside of the body could think and compute, as we spent more time exploring the capabilities of these neurons, it soon dawned on us that we might be on the cusp of a new paradigm of computing and thus intelligence. We discovered that these hybrid biological chips were capable of self-organizing and restructuring to solve novel problems that they had never encountered before without explicit wiring or programming by an engineer or programmer.

Instead, we believe that their malleability for independently solving problems is a genetic endowment shaped by millions of years of evolution. For instance, our experiments have shown that neurons can learn how to play the game of Pong and how to improve its performance over time without explicit human supervision. Our findings were recently published in a preprint manuscript here.

Although casual observation of this brain-computer hybrid that we have named dishbrain, appears to have mediocre game playing performance, the important thing is that it does hit it more often than expected by chance. Importantly, it shows an improvement of play performance over time that is statistically significant which confirms our hypothesis that the chips are indeed learning.

A more exciting observation that we have made is through our second simulation environment which replicates the Jumping Dinosaur game when the Chrome browser is used offline. We have noted that neural cultures that have learned how to play the Pong game also exhibit the ability to learn how to jump the dinosaur over obstacles simply by telling the cultures that a successful jump over the obstacle is good while a collision is bad.

The implications of these findings are profound: that these could represent a novel computing/intelligence paradigm that learns through experience and is able to re-wire and re-program itself without explicit human instructions.

What separates our work from that of other researchers is that we have been able to shape behavior in the hybrid chips and demonstrate its capability in a robust and reproducible way.

How we shape these behaviors is by having the neurons embedded onto the surface of a multi-electric array device that has microscopic sensors that read when the neurons discharge an electrical charge (for example, to signal when to move the paddle), and then stimulates them with a tiny electrical charge in return (to signal when the ball has been hit). The key achievement has been the custom low-latency software developed that enables an action-perception simulation loop to complete quickly enough that the neurons can associate their action with feedback.

The fact that the neurons have been able to learn how to play the game is what the team hoped for. But the circumstances under which it can do so has been an eye-opener.

In one of the experiments, we decided to shut off all feedback when the ball was missed, briefly plunging the neurons into a dark sensory-deprived world. This lack of information provoked a response, and the chips actually learned better as a result of it.

Although further studies need to be done, the team was excited by the findings. As our Chief Scientific Officer, Brett Kagan says: “We are seeing some very early evidence that these simple cultures are displaying signs of apparent ‘curiosity’.”

One thing we’ve found is that the chips made with human neurons outperform those with mouse neurons. As far as we can tell, this is the first research to provide empirical evidence that human neurons may be better at processing information than rodent ones due to their cell structure alone.

Our hope is that this self-organizing ability to solve problems in unfamiliar situations is what will make these chips useful. Programmers could simply simulate environments, set tasks, and let hybrid-biological computers determine the best ways to achieve them. This is a paradigm shift in the way we think of computing and how we build applications for this platform.

There are other advantages to a biological computer chip. For example, they are more robust to physical damage, are able to adapt or reorganize to change; they are scalable without generating too much heat, able to grow neurons without the need for expensive absolute-zero facilities or nano-scale fabrication units; and they will consume far less energy than conventional computing (the billion neurons in the human brain works on only 20 watts of power).

Given the broad range of potential applications, there is an opportunity to collaborate with other parties. Potentially, this research field will be so big that no one company can fully understand and exploit the technology to its true limits. It will require an entire community of like-minded and curious researchers to bring to life the possibilities of such technology.


  1. The chips are immersed in a life-sustaining liquid medium on which the neurons are placed on top of a microelectrode array containing about 22,000 sensors that transmit and receive electrical signals to and from the neurons.
  2. Cortical Labs’ whitepaper is currently under peer review.

Hon Weng Chong is the Founder and Chief Executive Officer of Cortical Labs. A startup headquartered in Singapore, Cortical Labs has raised around S$2.2 million (~$1.61 million) from Blackbird Ventures, January Capital, Westcott Family Office, and two individuals, Miles Albert and Vishal Garg.

TechNode Global Insider publishes contributions relevant to entrepreneurship and innovation. You may submit your own original or published contributions subject to editorial discretion.