What else has artificial intelligence a hell of a head start? The human brain. And exactly: researchers want to use human brain cells in biocomputers to design the AIs of tomorrow.
Ever since the advent of ChatGPT, we’ve been talking about it all the time. Artificial intelligence is now ubiquitous. She can code, write essays, draw, make music, give speeches and even write books. But AI has also shown its limitations. To go even further and make sure we better understand what’s in our heads, scientists at Johns Hopkins University in the United States added a bit of real “brain juice.”
They have developed what they call an organoid intelligence or IO. It might be scary, but there really are human brain cells in this IO. The organoids in question would be collections of living tissue derived from stem cells. They would be nourished with fluids through microvasculature. Their electrical signals would then be picked up by microelectrodes. Finally, the result would be analyzed by algorithms that rely on AIs as we know them. By comparing several of these “brain pieces,” the researchers envision that performance might ultimately be similar to that of the human brain. This biocomputer with 3D cell structures is described in an article published in the journal Frontiers in Science. The lab has even developed rudimentary prototypes of IO. They are not even vascularized and remain very fragile.
A biocomputer with low consumption and colossal performance
To communicate directly with this organoid, the team developed a brain-computer interface inspired by EEG (electroencephalogram) helmets. Developed last year, this flexible sheath that encases the neural matrix uses an incredible density of tiny electrodes to pick up signals from the organoid and see what it’s “thinking”. For now, this pseudo-brain remains far from the performance of ChatGPT given its small size. It only has 50,000 cells, which is quite an achievement. But it would take more than 10 million or 125,000 billion synapses for the IO to pulverize the current AIs. Each of these synapses could store up to 4.7 bits of information. Better than the world’s most powerful supercomputer with its 58 billion transistors, each capable of storing 4 bits.
However, as the publication explains, this IO would have one major advantage, which is low power consumption. It could approach that of a human brain, which at 3 pounds consumes just 20 watts for roughly 1 exaflops of computing power. On the machine side, this record was only achieved by a supercomputer last summer. But it took phenomenal energy to pull it off, not to mention the colossal dimensions of it all. By combining the best of both worlds, the lab believes that learning the IO would be tremendously easier compared to a traditional machine, in addition to this reduced energy consumption. But right now, current AIs have plenty of time to carry out their revolution. According to the researchers, these surgeries should start in ten years at the earliest.