defining the limits of whole brain emulation

The quest to create an accurate simulation of the brain continues. But what will be this project's most realistic outcome?

Here’s a quick WoWT trivia question for you. Who is Henry Markram and when was the last time he appeared on this blog? Did you say he was a neuroscientist working on an IBM project to accurately map the brain, and that the last time he was mentioned was in my review of the now famously overblown cat brain simulation claim by Dharmendra Modha’s Blue Matter team and my interview with him for Discovery Tech? Say, how did you… Oh, yeah, the search box. Right. At any rate, Dr. Markram is still trying to get an accurate simulation of a fully functioning brain up and running to understand some very fundamental questions about how our neural circuitry comes together. To give you an idea of his lofty goals, here’s his TED presentation at Oxford in which he provides an overview of the subjective nature of reality and perception, and how to go about replicating the relationship between perception and decision making and the mechanisms that enable it on a computer…

Overall, a very interesting presentation, but there are a few quibbles that have to be raised. I’m not really sure if Markram’s goal was to give the audience a brief rundown of how the brain forms a picture of its surroundings, or if he’s actually saying that we all share different realities built up almost exclusively by our brains, but this is actually kind of a misnomer that’s easy to take the wrong way. New Age woo-meisters like Chopra have been abusing this idea for years, declaring that the universe is just an illusion made up by conscious decisions on the part of humans and by liberally applying the word quantum, we could somehow master this reality. But the truth of the matter is more complex than that. Your brain doesn’t scan and absorb every last detail of what you see. Instead, it absorbs whatever’s relevant to navigating your environment and anything above and beyond is usually either ignored, or glossed over by your brain’s remarkable powers of inference. Optical illusions often exploit this mechanism to surprise you, or just play tricks with your ability to sort the patterns your eyes see as they’re refined through the V1 to V5 areas of our brain.

Meanwhile, the reality you see is still very much there. You just interpret it differently than another person since you have different patterns of neural connections in your cortical columns. Just because you missed a detail, or noticed another one, doesn’t mean that a particular object doesn’t exist or was willed into being by any other human. And if we build up on the same concept and look into abstract reasoning, we’ll find that it’s a massive and unavoidably unique compendium of experiences and memories that builds up your opinions, personality, and worldviews. Your cognitive biases accumulated from the culture in which you grew up, your education, the people who influenced you by something they said, etc. And because everybody’s neurons are different, you’re going to end up with a personal, unique circuitry which works the same way as that of other humans (sends a series of electrochemical signals at certain frequencies and patterns through a neural circuit), but it wouldn’t be interchangeable with any other set of neurons. And this is to be expected. After all, evolution doesn’t exactly standardize living things because if it would, its rate would be far slower and it’s unlikely that any multicellular organisms would’ve evolved because they couldn’t accumulate enough mutations to do so.

Likewise, this is why I kept pointing out why simulating a brain for an AI could take a human lifetime. If you’d successfully simulate a human brain in a supercomputer, you’d have a unique individual setup, not a general blank slate on which you could experiment at will. This is why Singularitarians who believe that you could just upload your mind with future super-technology are grossly mistaken that whole brain emulation would be a step towards this. Every brain will need its own simulation and simply creating an AI functionally similar to the human brain wouldn’t help either because the AI would use standard, immobile circuitry rather than a unique, highly plastic setup of an organic brain. In fact, the further you get away from accurate biological simulations of every neuron specific to an individual, the further you step away from even the most remote possibility of mind uploading. Whole brain emulation will be great for trying to get an idea of how the brain may work based on an accurate replication of our best knowledge about neurology on a supercomputer. But that’s all it will be. We’re not going to learn how to construct new universes by conscious thought, or how to port our minds over to new and different substrates. The biology and mechanics of our cognition just don’t work that way.

# science // computer science / neurology / scientific research


  Show Comments