One Neuron Can Be as Complex as a Large Deep Neural Model

This piece, by Onno Berkan, was published on 09/24/24. The original article, by Allison Whitten, was published in Quanta Magazine.

Scientists have long been interested in understanding the true computational complexity of the human brain. This is in fact what led to the creation of artificial intelligence in the first place– an itch to understand what goes on in our own heads. 

With today’s AI systems being bigger and stronger than ever, the issue of efficiency– the fact that the brain can process information much more effectively than any computer out there– has led to researchers going back to the brain. One such study investigated the computational complexity of a single rat cortical pyramidal cell and found that it takes a 5-8 layer deep neural network to model the activity of the neuron.

This brought into question the common analogy of the brain as a large computer. If it takes 5-8 layers and over a thousand artificial neurons to simulate just one biological neuron, is it really fair to compare the two? Well, it still may be, as the basic principles of the artificial and the biological neuron remain the same. Both are, at their most basic, computers. The issue is that the biological neuron is further than that most basic than researchers had previously thought.

The researchers ended up creating a massive input-output model of one of the most complex neurons in the mammalian brain, fed the information into a deep neural network, and reached 99% accuracy at the millisecond level. This doesn’t necessarily mean that the neuron and the neural network are equally complex, however, as the neuron is a part of a larger system and thus its activity reflects the complexity of that system. That is to say, you’re not really modeling a single neuron, you’re modeling a zoomed-in system.

That being said, the researchers were unable to “record the full input-output function of a real neuron”, which suggests that biological neural networks may be way more complex than 5-8 layer artificial ones. As the study’s author said, “We’re not sure that between five and eight is really the final number.”

Want to submit a piece? Or trying to write a piece and struggling? Check out the guides here!

Thank you for reading. Reminder: Byte Sized is open to everyone! Feel free to submit your piece. Please read the guides first though.

All submissions to berkan@usc.edu with the header “Byte Sized Submission” in Word Doc format please. Thank you!