Recently, an Australian team from Cortical Labs achieved an impressive milestone — living human neurons learned to play Doom and demonstrated reactions faster than GPT-4.
The essence of their experiment is as follows: scientists grew neurons directly from skin and blood cells of adult donors using reprogramming techniques. After that, they connected these neurons to a specialized biocomputer called CL1, designed specifically for such purposes.
The result was astonishing — these living cells not only managed the task but also outperformed one of the most advanced language models, GPT-4, in processing speed. A smaller group of living tissue adapted more quickly than a large neural network.
It’s interesting to note that a similar Pong game experiment conducted by the same team earlier took over a year. Now, they have taken a step forward — they have opened cloud access to CL1, are actively developing SDK software, and plan to release their biocomputer as a commercial device.
The main idea is to use this platform for medical applications: modeling neurodegenerative diseases, testing new drugs, or studying mechanisms of neural recovery.
However, important questions arise. For example, neurons themselves do not possess consciousness — as believed by Cortical Labs — but they originate from living humans and exhibit behavior very similar to learning. Where is the boundary between just a biological model and something more complex? And what should be done when systems become even more advanced? No one has definitively defined these boundaries from legal or ethical perspectives yet.
These are only the first steps in a highly promising direction that is already shaping the future development of technology.
Created with n8n:
https://cutt.ly/n8n
Created with syllaby:
https://cutt.ly/syllaby