Human Neurons Still Faster Than AI

An Australian company called Cortical Labs has achieved something incredible. They have created a miniature brain called "DishBrain," which is nothing more than a mini-brain created "in vitro" by connecting thousands of human neurons, which they stimulate electrically with a network of microelectrodes.

In addition to this novelty, what is interesting is that the scientists at Cortical Labs have taught this mini-brain to play the classic video game called "Pong" in a matter of only 5 minutes, much faster than the 90 minutes needed by the most advanced AI that exists to date.

Although studies had already been done with neurons "in vitro," this is the first time these neurons have been able to perform specific tasks.

But How Was This Possible?

The first thing the researchers did was create a simplified version of "Pong" for a single player. Then, through electrical signals to the right or left of the "DishBrain," they began to teach the neurons where the ball was located in the game.

Once the mini-brain learned this, it began to move the racquet back and forth to hit the ball on its own.

What is particularly interesting about this study is that human neurons could learn much faster than AI.

One possible explanation for this difference is that AI uses methods such as machine learning and deep learning, which require a "massive" number of computations to adjust the parameters of their algorithms, to which the point of requiring long periods of training.

On the other hand, human beings learn very quickly because our brains are able to make use of a priori knowledge, something that is innate and that allows us to "filter" the information we receive.

What the scientists achieved, in essence, was to trick the mini-brain into believing that it was the racket in the virtual space, as Brett Kagan, scientific director of Cortical Labs, told to NewScientits.

As explained in an article published in the journal Biorxiv and is awaiting review and contrast by scientific peers, the mini-brain can feel and, on its own, learns to perceive its virtual environment and develop within it.

Thanks to "DishBrain," it is being demonstrated that a layer of cortical neurons cultured "in vitro" can self-organize and display responsive and intelligent behavior within a simulated game world.

The Brain of Future Cyborgs

What has undoubtedly shocked scientists the most is how quickly biology has been able to overtake AI. While it currently takes 90 minutes for AI to learn this game, only 5 minutes of stimuli were enough to generate the necessary learning for this biological mini-brain.

However, despite the speed of learning, scientists have noted that both humans and AI are better at playing "Pong" than the "DishBrain."

Still, this discovery will undoubtedly allow the integration of these biological neurons within the power of AI to develop complex brains that will govern machines in the near future.

As is always the case, what was only science fiction a few years ago is getting closer and closer to becoming a reality. Machines will soon have powerful brains that will make them smarter and more efficient than humans. We will see what happens when these kinds of discoveries reach their maximum level of development.