# How realistic are my neurons?

I am writing a book about anatomically plausible superhumans ( a new artificially created form of man ), which, in addition to incredible physical strength, should also be distinguished by incredible intelligence, which I have always tried to increase simply by making their brains larger or compacting the neurons themselves, as is the case in birds. However, some time ago I thought about why not transfer some of the functions into the cell itself, creating a primitive computer from each neuron?

So in binary computers we can only pass or delay one signal, depending from the other. We use Boolean algebra and logic for addition. At the same time, an analog computer can fold natively. Gapless synapses (a design of neurons invented by me that might not have a gap between synapses) already know how to amplify or jam a signal, and we just need to add signals to each other and change their total power according to a given coefficient. How should such modified neurons be arranged?

Unlike ordinary human neurons, inside this neuron there should be thin branching tubes, divided into sectors. Sectors of these villi could connect, amplify and separate the signal, form direct and feedback stations, and pass the impulse in only one direction. In this case, the tubes themselves would have to be filled with chemically obtained ions, which would regulate the electrical resistance of the medium. Depending on the potential of the pulse, the conductivity can change, which changes the power of the signal at the output. The signal would have to be added if the two tubes are connected together. The signal that passes more often stimulates an increase in the channel conductance, which amplifies it at the output. Thanks to external signals, it is possible to change the conductance of individual channels, which makes it possible to change the weight of the entire neuron. As a result, such cells should have become larger, but a network of ten such analog neurons could replace a network of a thousand ordinary ones.

How scientifically realistic is this design of neurons of my superhumans and how much, if it is possible to calculate, ordinary people could become smarter with such an addition in the nervous system?

• I'm a bit confused about your goal. It sounds to me like you want the function of synapses, where signals can be amplified or reduced depending on the state of the body and how often they are used, moved to these 'computer neurons'. I'm a bit at a loss why and how you then send the signal to the correct neurons after the signal modification. Jul 14, 2021 at 12:59
• (1) Analog computers are slower than digital computers, because of how analog cirtcuits work. (2) Not clear what you mean by "signal". If you mean an electrical signal. i.e., voltage, then no, no way, you cannot make a logical circuit by simply splicing wires together. It really doesn't work that way. (3) A resistor can only reduce the amplitude of a signal, it cannot increase it. (4) I've never heard of electrical signals represented as power -- either voltage or current, never their product. (5) The question is using the words signal, power and potential as if they were synonymous. Jul 14, 2021 at 13:11
• P.S. Real-life neurons actually are active "computing" devices. They can act as adders (summing up the inputs) or as integrators (firing when the inputs in time add to a certain threshold). (Of course, they are not truly computing devices -- the data is represented on the nerves in a quite complicated way, depending on the specific nerve. It's not really straightforward analog. Look for example how data is represented on the auditory and visual nerves -- it's really sophisticated.) Jul 14, 2021 at 13:12
• I'm sorry for the inaccuracies in the use of some words, I just don't know English, and the translator that I use does not do its job perfectly. Jul 14, 2021 at 13:26
• Please pick one tag out of reality-check and hard-science, these are mutually exclusive. I suggest that you probably want reality-check, since it would be very hard, maybe impossible, to comply with the hard science requirements for a neuron model that you have made up. Jul 14, 2021 at 13:48

I've read of many fringe science (and let's be honest, most of these are rather deep into pseudoscience) that claim various things about neurons:

1. That they aren't the minimum unit of computation in the brain, that there are specialized organelles that perform computation and neurons are a primitive network of computers.
2. That signaling between them isn't electrochemical, but "acoustic". Supposedly that explains the rather low signal times. Myelin degenerative disorders are, allegedly, explicable because it interferes with that signalling.
3. That neurons aren't even the seat of computation in human minds, but that their activity is merely some side effect.

And so on.

Much is unknown about how human intelligence works. There is quite a bit of wiggle room, especially if you're willing to dip your toes into the pseudoscience.

However, it should be pointed out that engineering these things requires an even more sophisticated engineer who understands it all, presumably who isn't superhuman himself. Further, unless humans already have sub-neuron computation/intelligence, then actually engineering that as a feature of some GMO is so far out there that it's unlikely anyone today or for many millennia could accomplish it.

As for calculating intelligence, we can't even measure it meaningfully. Go read about the Flynn Effect.

• In other words, do you no know if my voiced idea of modifying the body of a neuron is realistic so that it can work as a primitive computer, as you correctly noted, having smaller structures inside for this? Jul 14, 2021 at 14:34
• @FrankThompson4 Computation itself is trivial, you can make basic computing elements out of wooden gears. It is not in dispute that a biological cell could include elements/organelles that have computational capacity. Whether this would even be useful to a human-like intelligence is another thing entirely... human intelligence doesn't use Turing computation, but neural network computation. NNs are simulatable on Turing devices, but your cells might not add much in that regard. Jul 14, 2021 at 15:19
• Nevertheless, the main question concerns exactly whether it is possible to create the system described by me? Roughly speaking, by transferring a neural network ( we are talking about the brain, and not about self-learning networks from computers ) from the tissue level to the cellular level. Jul 14, 2021 at 16:01
• If it still cannot be done or it turns out to be not cost-effective ( it will not increase the total computing power of the nervous system of my superhumans ), then what options could be used instead? Jul 14, 2021 at 16:04

and we just need to add signals to each other and change their total power according to a given coefficient

Yep. Why not just do that.

This suggests to me that there's a fairly pseudoscientific lack of appreciation about just how meshed and intertwined and jawdroppingly complex, neurology is. "Just" change their "power" according to "some coefficient". Hmmmm.

Based on this, and other science sounding word salad, I'm honestly not hopeful that this would be at all "realistic".

That said, its your world. If you say it works there, then by definition there isn't proof it can't, either.

This question appears to mistakenly assume that neural networks are digital computing devices. They are not... they are analog. Each neuron acts as an integrator, summing its inputs according to how close each is to the body of the neuron, and how temporally closely those inputs arrive. If the sum of the inputs exceeds a threshold, the neuron fires... though it may fire at a baseline rate without input.

So, to so radically change the way that neurons work as the OP describes, the likelihood is that the resultant brain would either fail to function, or function sufficiently differently that its thought patterns would be considered alien.

Intelligence is a factor of processing power applied over time. It is possible for a person with a lower effective processing power to be perceived as more intelligent by virtue of spending more time thinking. There is also apparently a thought-time limit to most people's brains... people tend to spend a fixed time thinking about an abstract problem before concluding their thoughts and either announcing the answer they have come up with, or admitting defeat, but this time limit varies from person to person.

So, the best way to increase intelligence is not necessarily to radically change the way that the brain functions, but to remove the inefficiencies... reduce the propagation time of neural impulses and to use faster synaptic mechanisms than chemical neurotransmitters. That way, more processing can take place in a given time, increasing the intelligence of the person.