Monday, February 27, 2012

Global 2045: Randal Koene: Whole Brain Emulation Within Our Lifetime 1

We present here the text of a lecture delivered at the Global Future 2045 Conference in Moscow on February 17th 2012 by Dr. Randal Koene, of Halcyon Molecular.
Read In New Wide Format (or Internet Explorer Users)

Part of what I do is work at Halcyon Molecular to build the world's most accurate Genome sequencing, which is something that to do, we need people with lots of different skills.  My background for example is physics and electrical engineering, information theory, computational neuroscience and neural engineering.

Carbon Copies
But in addition to that, I'm the founder of a SIM group that I'll explain in a second, called Carbon Copies.  What I'm going to try to explain today is what that group does, which is to bring together the experts and projects needed to do something called substrate independent minds.  I want to tell you what that is, why it is important and how it can be done. Specifically I want to talk to you about how feasible it is.

In groups like this, we often talk about things like life extension, augmentation, matters like that.  When we do, I think it's important to consider what the objective really is.  What is it that we're trying to do?  What I mean is, what are we trying to extend or augment?  So I want to take a little step back and think what are we?  What are you?  If you look at these slides, you can see in the background a lot od different experiences a person can have.  So experiences are a part of you.

Another part of you is your body, the senses that you have, it's the actions that you can take, and then on the right you can see all these different expressions.  I'm trying to represent the idea that each one of us has unique characteristic responses to things.  So if you put us in the same situation, with the same input, a different individual reacts slightly differently.  There are unique responses that we have.

The Centrality Of The Mind
So where does all this actually takes place?  All of it takes place in the mind.  That's where we do all this stuff.  Really, I can't even touch anything in the universe.  That's how unreal it is.  Imagine that I'm trying to touch this cable.  If you think about the atoms in my finger and the atoms in the table, the protons and neutrons in there never actually collide.  It's mostly space.  There are some forces there.  There's electrical energy that goes up my arm and goes up in the brain.  I don't notice that until something is being processed.

So, when we talk about things like life extension, what we really mean is that we're trying to safeguard, those processes that being with experiences, that is going on inside our minds.  There are really two different ways that you can do it.

Two Different Approaches To Life Extension
One is that you can say, we know right now that this is processing is taking place in the brain. So what you need to do is you need to maintain the brain.  We need to make sure it keeps functioning well.  We need to make sure we keep the body functioning well so that it keep the brain going.  You need to make sure that your environment stays safe because  the body depends on that environment.

It's not built for every possible environment.  Imagine if we suddenly lost the atmosphere.  Well this body would die and the brain would go and everything would stop.  But we also just said that what we're really talking about is the processes going on in the mind.  That's the experience of being.  

So the alternative is that you could address that correctly.  Do what you would do, if you were treating it like a valuable piece of information, or a valuable program.  What would you do with a valuable program?  The first thing you would do is that you would make a backup.   You would archive something.  The next thing you might want to do is that you would want to be able to fix errors if they were there, or, you might want to change to run on the next best hardware out there.  So you want access to the code.  That's really what I'm talking about.

Substrate independent minds is about data acquisition.  It's about access.  The notion being that right now there processes in my mind are running on a kind of substrate.  But if they can run on many different kinds of substrates, then we call it substrate independent.  By analogy, think of computer programs that are platform-independent, that run on different kinds of platforms.

Which of these two strategies is better?  Well I think they are both valuable.  They should be looked at very concretely.  If you're thinking where would I put resources, where would I pur effort, what can we achieve within out lifetimes, then you need to look very concretely at the plans that are there.  So that's what I hope to do today.

But before I rush into that just a few quick terms.  What you hear a lot of people talking about in some communities when you talk about technologies like this is the term mind uploading.  Mind uploading is a perfectly good term but it doesn't say anything about what you do once you put your data somewhere, once you upload it.  So it's not very objective.  It's really about the track transition from this biological substrate to some other one.

Then where would you like to end up?  If you're trying to do substrate independent minds, there are many ways of doing it.  Preferably, if this were an ideal case, on every substrate you go to, the code would be optimized for that, like compiled for it.  But we do not really know it that well.  If you ask a neuroscientist do you know this part of the brain, they will say no.  Because you know that what you're actually asking is do you understand all the different strategies that the brain is using at different levels of the hierarchy right down to the neurons?  And of course we don't.  So we can't use that approach right now.  But what we can do is go down to the bottom of that level and we know more about how neurons work and how they pass particular information to one another.  This is what we call whole brain emulation.  The idea that you can emulate the neuro anatomy and the neural physiology at that level.

History Of Substrate Independent Mind Research
Now this is not an entirely new field.  I know it has not been talked about as much as maybe biological methods, like when you hear Aubrey DeGray talking about his foundation and things like that.  But there has been work going back to least 1994 by researchers dedicated to the pursuit.  There was the Mind Uploading Research Group that I inherited from Joe Straut who is the person in gray up there.  The Knife-Edge Scanning Microscope was developed by Bruce McCormick who's right next to him.  In 2007, we had an oxford workshop about whole brain emulation that road map came out published by Anders Sandberg, who you see in the picture that I am next to Suzanne Gildert and myself.  There are interesting people like Ted Burger and Ken Hayworth.  Sebastian Seung who popularized the idea of the Connectome, all the connections in the brain being important, at the 2008 annual meeting for the Society for Neuroscience.

We have the formation of Carbon Copies with Suzanne Gildert and myself.  Since then, it's become much more mainstream.  We have people like orthogenetics star Ed Boyden working on brain emulation and extracting circuitry from neural circuits function emulation.

Whole Brain Emulation
If you want to do substrate independent minds there are different routes you can take.  I'm only going to talk about one today.  Another really interesting one, besides the whole brain emulation is the brain computer interface route.  But it has a lot of other driving factors behind it.  It has a lot of commercial potential at this stage.  So it was so important that I thought, this year in 2012 I'll just do a whole separate workshop on that.

Click to enlarge
Leaving that for then, If you're going to try to do something like whole brain emulation, then you need to make representations of things.  Every time you make a model or representation,  you need to choose a resolution, some level at which you're going to describe the elements and describe what they do.  That means you need to characterize them.  When you're characterizing them at a resolution it means that below that level, below that resolution at any higher level, you don't talk about the mechanisms that might go on inside, instead what you're doing is you're looking at their input and their output, treating it like a black box and describing what the functions are there.  You want to capture all of them.  You don't want to miss any latent function.  In these boxes here are just some examples.  In the top you see how Wu et al for sensory neurons.  In the bottom you see the famous the Eugene Izhikevich neuron model, where he uses several parameters to easily model many different biological types of neurons.  On the right is some work of my own where I focus on currency of the brain which is the spikes in the system.  After all, when a neuron spikes, and the exact timing of that spike, is both is determined where memories are laid down and what things are remembered,  as well as which muscles are activated which interact with the environment, even just by speaking.  These are important issues if you want to model something correctly.

Then above that level, the important factor is to capture emergent properties.  You have these elements, you've characterized them,  and what you need to do, is you need to understand how they all work together, because this gives you emergent function.  A good way to find out what they're doing together is to first of all know how they're connected.  That's why extracting the human connectome with all the connections in the brain is really important.

Entire Person As A Black Box
Now you can do this black box selection, at which level you want to work on in many different ways.  In SIMs this happens in the routes people choose.  There's a route called loosely coupled offloading, or, at least that's what we call it generally speaking. where the entire body, a person is considered a black box, and what you're trying to do is model their behavior, model how they act so you could basically make a simulacrum of them and say this is that person.  What it means is that you need video recordings, audio recordings, life-logs, and maybe an AI that learns to how to interpret what you're doing.  This is very similar to a method called the Bainbridge-Rothblatt Model that is used for what they call trying to create an upload.  There are differing opinions about that.

Entire Brain As A Black Box
At the next level, you can take the entire brain or parts of the brain as the black box.  This happens for example when you find an interesting or correct architecture for the human brain, where you can say this area does that, that area does that, and you want to see how each one of them works and maybe personalize them so they seem like one person.

Neuron As A Black Box
At the next level, this is where it gets really interesting, you can take either neurons or parts of neurons, like the morphology of neurons, as the black boxes and this is what's done in computational neuroscience and neural informatics.  This is the level that I'm going to be talking about.  It's the one that is most concrete and most usable right now for whole brain emulation.  They're also interesting for brain computer interfaces of course.

Resolution, Validation, Connectome & Platform
Now if you're going to do whole brain emulation, you're going to look for it as a strategy, as a whole map.  There are really four big requirements.  There are four things you have to do:








  • You choose your resolution and scope as I just said.  You need to validate that.  You need to do some tests.  I know a few people who are, and it's very interesting work.  
  • You also need to extract that connectome, the structural connectome, where are the things connected as I was just saying. 
  •  And because you need to characterize all of those elements that you chose at that bottom resolution, throughout that connectome, basically you need to have a functional connectome.  And finally because you want to reimplement this,
  • you need some sort of platform that could emulate this
  • Just a quick word on function and structure, that's something that often gets brought up about neural networks, made to look rather mystical and of course, function and structure happens in every kind of system that we have.  If you think for example about a computer chip.  It's full of transistors, those are the basic elements and then they're arranged in a certain structure.

    [We will continue with part 2 of Dr. Koene's lecture in your next installment.]

    1 comment:

    Roeland said...

    I do not think mind uploading is a correct term. What one can do is make a copy of the mind and load that into a machine.
    However, the original mind remains as it is in the original body, unless someone destroys it.
    So the correct term would be copying a mind. I do not see the point in that because the life span of the original remains the same. So why do it?