Building each region of the brain & body’s entire interconnected system in software.
The connectome is a term first coined in 2005 and refers to the entire electrical system inside an animal’s body. It’s not just your brain, as some people mistakenly think, but rather your brain’s neural connections and how those spider out into your body’s nervous system.
Every animal has a connectome and when you hear the term neuron it, again, doesn’t refer to just the brain, but also all the connections in the body.
Consider any house you’ve ever been in. It’s wired for electricity within the walls and exposed with a wall outlet that you can plug things into. It’s the same around the world. Imagine for a moment that you could extract all that electrical wiring from one older house, then insert it into another newer house. As long as you connected the wiring to the outlets appropriately, the new house would function just like the old house. Plug something into the outlet in the wall and blamo, you’ve got power.
In this analogy, the house is your entire human body and the electrical wiring is the neural connections in your brain and nervous system.
We’ve created an artificial version in the house when the real biological version in the body is much more efficient.
Either way, you could consider both a connectome. Our scientists, however, have only managed to map one animal’s entire connectome. That’s a worm because it only has a few hundred connections. Others are in progress, like the Zebra Fish, and will continue to be mapped until we reach the ultimate goal of an entire human. The problem, though, is that humans have exponentially more neural connections throughout our bodies compared to simple organisms like worms or fish.
II. Mapping Brain Regions
Recently we have identified to quite a detailed degree, all the regions in the human brain. In many cases, we can multiply by 2 for the left hemisphere and right hemisphere but there are several regions that reside in one or the other.
Each of these regions is a “connectome” in and of itself. If you stop and ponder that for a moment you might be reminded of a mathematical concept called a fractal. In short hand, it means it looks the same whether you zoom in or zoom out.
Or as one of my research friends, Timothy Busbice’s case, an adjacency matrix that maps the neurons in that region to one another. That’s a fancy way of saying an “interpreter”. Not just for language, as us humans are used to thinking of them, but in terms of connections. This plug goes into that socket, kind of thing.
III. The AI Industry Problem
The big issue with connectomics is that most of the world still doesn’t know of its existence. Even in the artificial intelligence circles within Silicon Valley, it’s a term that most people aren’t familiar with. As the industry has moved from mechanical machines to mathematical software, many people focused on the deep neural network approaches are missing the biological next step.
Namely, the connectome. It’s a map of the connections inside any animal’s brain and body.
As you can see from the graphic we prepared below, there were no Google searches related to the term until mid 2005 when Olaf Sporns coined it.
I didn’t pick it up until 2011 when I started working on the AI inside StoryApp and began researching it myself. There’s a great background book by Sebastian Seung, a physicist turned neuroscientist, that I read back then. If you’re interested it’s a relatively easy, quick read.
Today, the study and interest in the connectome is overshadowed by discussions of AI and Deep Learning. You won’t see this term show up in the popular press, or even in some of the most popular tech press. I suggest the MIT Tech Review newsletter produced daily by Jamie Condliffe.
But I digress.
IV. New AI Research
Back to my friend Tim, who has developed the means to create the communication between these fractal-like neural connections. And he’s mimicing these biological connectomes using software. In essence, creating truly artificial biological intelligence (ABI?).
With artificial connectomes, the goal is to create regions that are either an expression of some sensory input or motor output, each with a potential for a number of cortical layers (e.g. the human visual cortex has 5 layers). We can then see if we gain intelligence from these configurations just as nature does the same. So yes, you could think of each region in the brain as a computer chip.
If I may quote an email exchange we’ve had, the following is directly from him:
The granularity would be dependent on the complexity of that region or said better, the number of neurons involved. If we have 1000 neurons, we map these neurons into an adjacency matrix of 1000 x 1000. I believe neurochip designs are very similar. So if there are a 100 billion neurons in the human brain and approx 180 brain regions, we are talking approx 555,555,555 neurons per region, or 500 Million per region. Sounds like a lot and the issue really does become having an entire connectome of 500M processing simultaneously. However, I suspect that each of these 500M can be broken into much smaller subregions to make it more manageable. I can see this as a possibility that is achievable because of my tech.
Even if we have 100 Billion neurons in our brains, there are several aspects of the brain we can eliminate for robot use. Obviously, our robot doesn’t need to drink and pee afterwards, and there are a hefty number of neurons used to regulate and control our bodily functions including breathing, heart beat, food absorption, etc. Most, if not all of these neurons can be removed and this is one of the main arguments for artificial connectomes = we need only apply brain regional/neurons to the aspects of robot operation and intelligence, not things that are not robot specific. The actual number of needed neurons is unknown at this time but I can certainly hypothesize that it is smaller than 100 Billion.
My favorite part? Imagining a peeing robot.
V. Robotic Threats
There is a potentially chilling aspect of our research, however. It’s not all candy canes and unicorns. An intelligent robot could be as comfortable “living” on the moon as it does on earth. Thus, the things we humans hold dear such as nature, clean air, water, and food would not necessarily be in the best interest of an intelligent robot.
In fact, things like water may be a threat and thus push an intelligent system to want to rid the planet of this threat. Living inorganic beings would most likely have a different sense of whats good and bad, and something we need to be very cognizant of as we delve into this new world.
Ponder that over your next pumpkin spice latte. Because robots don’t much care about Starbucks.