Image on left by the PIT Bioinformatics Group. Image on right by the OLCF at ORNL.
Creative commons licensing. Click for specifics.
I have been working on my new book, Science in the Atomic Age, which (Lord willing) will be published this summer. In the section where I cover the nervous system, I compare a mouse brain and a human brain to computers. It’s rather fascinating. Below, you will find a slightly-edited excerpt from that discussion. Please note that the students have already learned that neurons are cells found in nervous tissue and that the integumentary system is the system of organs that makes your skin:
The brain has three major divisions: the cerebrum (suh ree’ brum), the cerebellum (sehr’ uh bell’ uhm), and the brain stem. The cerebrum is in charge of most of the really complicated things that the brain does. For example, it receives signals from your eyes and interprets them so that you can see. It receives signals from your ears and interprets them so you can hear. It receives signals from all the nervous tissue in your integumentary system so that you can figure out what you are touching as well as things like whether you are too warm, too cold, or comfortable. It also helps you learn, and it stores your memories. All this takes a lot of work, so it requires a lot of neurons.
How many neurons? The average adult cerebrum contains about 20 billion neurons. That number doesn’t mean very much by itself, so by comparison, the average adult mouse cerebrum contains about 2.5 million neurons. So the human cerebrum contains about 10,000 times as many neurons as a mouse’s cerebrum. Of course, a mouse is much smaller than a person. By weight, a person is about 3,000 times as heavy as a mouse. At least part of the difference between a mouse’s cerebrum and a person’s cerebrum is due to that. But people are much more intelligent than mice, and the number of neurons in the cerebrum must also be related to that.
So let’s look at another comparison. In 2014, a study tried to compare the complexity of the brain to that of a computer. The study concluded that a personal computer could simulate the processes that are known to happen in a mouse’s cerebrum, but it wasn’t nearly as fast as a mouse’s cerebrum. The personal computer ran those processes about 9,000 times slower than the mouse’s cerebrum. In addition, remember that neurons need energy to stay alive and do their job, while computers need energy (in the form of electricity) to do their job. The personal computer simulating a mouse’s cerebrum was not only a lot slower than the mouse’s cerebrum, but it also consumed 40,000 times more power! In other words, a personal computer is very slow and inefficient compared to the cerebrum of a mouse.
So what about a human cerebrum? How does it compare to a computer? Right now, there is no computer in the world that could even attempt to simulate what we know goes on in a human cerebrum. The world’s most powerful computer is the Summit supercomputer, which is pictured above on the right. It has a million times the computing power of the best personal computers, but it is still ten times too weak to simulate what we know goes on in the human cerebrum. The study I have been talking about indicates that even if a computer capable of simulating a human cerebrum gets built, it will require 42 million times the power that the average human cerebrum requires. So our very best computers are primitive compared to the human brain!
Now before we continue this discussion, I need to make two points. The first one is obvious. Since our very best computers can’t come close to simulating what we know happens in a human cerebrum, it is rather obvious that the human cerebrum is very well designed. It should be, since it was made by the ultimate designer, God. But let’s consider the mouse’s cerebrum. Even though we can simulate what we know happens in a mouse’s cerebrum with a good personal computer, the computer is still very slow and inefficient compared to the mouse’s brain. Once again, that’s because the ultimate Designer, God, makes better things than we can.
The second point is also very important. Did you notice how I kept saying that computers can “simulate what we know goes on in a” cerebrum. The “what we know goes on” is very, very important. We don’t know everything that goes on in a cerebrum, and even if we did, we probably wouldn’t understand it. We know a lot about brains, but most scientists who study the brain will tell you that there is probably a lot more that we don’t know! So even though a personal computer can slowly and inefficiently simulate what we know goes on in a mouse’s cerebrum, it can’t “think” like a mouse. That would take a level of understanding that we haven’t come anywhere close to achieving. In the same way, even if there comes a time when a supercomputer can slowly and inefficiently simulate what we know goes on in a human cerebrum, it will not be able to think like a human, because we don’t understand anywhere close to what we would need to understand in order to get that job done!
And the human brain runs on a miniscule 20-watts of power! Research concluded that if we could build a CMOS computer that had approximately the same amount of computing power as the human brain, that conventional computer would require a 10-megawatt powerplant to power it!
I love to write. Let’s make that plain so you understand why I’m a little weird 🙂 I love to write. Hard SF and space operas, and any computer without a human brain is referred to as an ‘idiot cousin’ by the cyborg. No one ever developed a better computer than God has done for us. You walk in beauty
I really enjoyed this comparison Dr. Wile. I am a Junior studying Software Engineering and I find it entertaining how much faith people are putting into machine learning. Not only can we not build a computer as powerful as a human brain, but our most advanced idea of how to teach a computer to learn requires an enormous training set. Teaching a human to recognize the letter ‘a’ requires only a couple of examples and they are good for life. Teaching a computer the same thing necessitates thousands upon thousands of examples for it to “learn” off of.
Thank you for the very interesting article Dr. Wile.
For those who haven’t seen this video previously, I will link a very short video of a small portion of nerve tissue from the mouse retina. Like the brain, the retina is composed of nerve cells or neurons. According to information given in the drop down box and the linked additional information, the block of tissue in the very short video I will link contains 950 neurons and about half a million neuron connections (synapses). The video illustrates the connectome, which is essentially a wiring diagram for the brain.
A very simple way to think of the connectome is to think of the wiring diagram for your home, cell phone or computer. It is essentially a schematic of the “wiring” of the brain. The connectome only represents the way the neurons or “wires” of your brain are connected. The underlying neurochemistry within the neurons and the synapses is much more complex than the connectome. A single neuron can have 10,000 or more synapses and each synapse is essentially a microprocessor of intracellular information.
The important thing to remember about this video is that the amount of tissue illustrated in the video is only 0.1mm on each side. That is approximately the same dimension as the thickness (not the length or width) of a dollar bill. If my calculations are correct, flying through the entire human brain at the same speed as the video representation of the neurons in the mouse retina would take over 100 years. I base that on 86 billion ( more recent research figures than the commonly quoted 100 billion) as the average number of neurons in the adult human brain.
https://www.youtube.com/watch?v=_n4YjsYN5sQ