Science is once again following in the footsteps of science fiction as IBM this week announced plans to develop, in partnership with Switzerland's Ecole Polytechnique Federale de Lausanne's (EPFL) Brain and Mind Institute, a working computer simulation that emulates the behavior of the human brain down to the cellular level. The project is called Blue Brain and will make use of four Blue Gene super computer racks, currently the fastest computers in the world with a peak performance of 22.8 teraflops. The result will be a 3-D working model of a human brain in action. The initial phase will focus on the Neocortex, the outer layers of the brain associated with higher functions.
This could well represent the first step down the path towards a "scruffy" Artifical Intelligence. Scruffy (or connectionist) AI represents the view of intelligence as an emergent quality, as opposed to the classical or "neat" path, which attempts to build intelligence with a more top-down approach. The thought is that this might prove to be a way around the problem of how to get the Ghost into the Machine and all the attendant philosophical and logical conundrums associated with that goal. Skirting the question of what exactly the Ghost is, the Artifical Life approach exemplified by Blue Brain proposes to create a machine that has awareness by emulating the behavior of a system that is already presumed to have awareness (i.e. the human brain.)
Fans of William Gibson's Neuromancer may recognise in this one of his favorite themes, which I like to call the "soul in a box": Expert systems which are created by taking snapshots of real people's brains and then allowing these simulations to continue to act as agents in the virtual world. The wrinkle is that these simulations, besides having all the expertise of the original "donors," also possess a sense of self identical to the donor's. To the artificial being thus generated, it as though it still were that person, and their organic life came to an end the moment they stepped into the MRI.
I have to admit I've always loved this idea. Especially if it ever becomes standard operating procedure. Imagine if instead of having "Clippy" the talking paper-clip office assistant pop up on the screen to answer questions or offer suggestions, you could have one of these virtual experts instead? Somebody really smart, like say Stephen Hawking? Wouldn't that be cool?
"Yes?"
"Hi, Mr. Hawking. I was just having trouble doing this mail merge and I was wondering if you could give me a hand..."
"That's what you're asking me? How to do a mail merge? You're sure you don't want to know something about advanced theoretical physics?"
"I really need to know how to do a mail merge. It's kind of important."
"You have access to all the knowledge and expertise of one of the most brilliant minds of the twenty-first century and that's what you're asking me!?"
"Well, uh, yeah. Look, maybe this is a bad time. Would it be better if I talked to Clippy?"
"No! Don't go! Please, I get so bored. I'll tell you how to do a mail merge if that's what you want. Hold on, I have to read the manual."
---------------------------------------------------------------------------------------------
Also in brain news this week: Researchers believe they have isolated the brain region responsible for recognising sarcasm and irony. And the winner is: The right ventromedial prefrontal cortex! Huzzah!
It is in the right ventromedial prefrontal cortex that the brain compares the literal meaning of the words (which is evaluated elsewhere in the brain) with the social and emotional context in which those words were spoken. Subjects with damage to this area of the brain showed an inability to detect sarcasm.
It remains unclear if the combination of this research with the above-mentioned brain emulation research will ever lead to the development of a machine which can appreciate the Simpsons on more levels than I do, but it certainly looks to be a promising field of inquiry.
---------------------------------------------------------------------------------------------
And finally, a story that brings us full circle, from trying to cram the unruly, chaotic contents of the outside world into the smooth, perfect confines of the box to trying to bring the contents of that box out into the real world: This BBC story tells how researchers in Singapore are developing a real-world Pac-Man game. Players wear virtual reality goggles as they navigate an actual maze. The goggles show them where cookies are and tell them when the ghosts are blinking, etc. Ghosts are also played by real participants. Power-ups are represented by physical sugar jars equipped with Blue Tooth technology.
Real-world Pac-man! It's like a beautiful dream...
Posted by flamingbanjo at June 6, 2005 09:17 AMi was getting really into the idea of the Blue Brain, but then you threw the real world pac man at me and i was truly impressed and excited. i'm ashamed to admit how fun that sounds. if they had that at the gym, i'd get a lot more exercise.
back to the blue brain .... I, Robot, which a pretty cheesy movie, really brought up a lot of the sort of ethical and theoretical questions around this (while barely addressing the technology), and i have to say that i'm really on the fence. i have no idea how i feel about it for many reasons, including that it sort of makes me feel .... empty to know that my brain and my self could be recreated digitally. i think more than any of the other aspects, that's what really keeps me from being 100% into the idea. somewhere in there it feels like the way humans define themselves will be seriously disturbed, and perhaps not in a good way.
Posted by: amy.leblanc at June 7, 2005 01:26 PMReal-world Frogger showed initial promise, but the liability issues proved cost-prohibitive.
Posted by: flamingbanjo at June 7, 2005 02:00 PMYeah, and real-world Burger Time just makes you fat and lazy...
Posted by: KING COMTE I at June 7, 2005 02:14 PMWeirdly enough, Battlestar Galactica does some interesting things with the theme of Artificial Intelligence.
Posted by: Joshua at June 8, 2005 04:12 PMI've been toying with ideas like the Blue Brain concept for a number of years, and I like the basic concept ("behavior" emerging from a system that is built to model low-level mechanical interactions). I think it's the only way that makes sense for life and intelligence to come into being, in my atheistic world view. A brain, however, is an exceptionally complicated set of equipment, and it requires a whole lot of knowledge we don't have yet to map it on such a level of detail.
A couple of years ago, I discussed with my father (a neurochemist with 40+ years of professional research experience) the idea of creating something a little simpler as a cellular model, with the idea that a computer model of cell behavior would be needed first. He suggested I try a nematode - a tiny multi-cellular organism whose complete life cycle is already understood in pretty thorough detail, and actually mapped on a cell by cell basis.
On more reflection, even that was kind of intimidating. There is so much more that goes on in an organism than just modelling the cells. For instance, you have to model the medium in which the nematode lives in order for it to get nutrients, and eliminate waste. If its "living in a vacuum", it would just "die".
Bringing it back to the brain, I personally believe that the whole body is an interconnected system, and the brain cannot be mapped in isolation. How can you have accurate brain behavior without simulating (incredibly complicated) input from the eyes, ears, and other nervous receptors that are placed throughout the body? A "Brain in a Box" can never behave like a human being, because it will never have the stimuli of a human being. It won't get hungry, or slam its fingers in the car door, or feel the loving relief of a full-body embrace.
Even if we could develop a complete set of inputs to simulate these events, they would occur at the whim of someone who pushes a button (or writes a program to feed them in), rather than occurring in response to real or perceived threats with genuine costs and consequences. Therefore, any "learning" in the artificial brain would be very contrived, and the lessons learned would quite likely be very disordered.
I think it's a fascinating idea for a project. I also think it is over-ambitious to our current level of understanding of the biology of the human brain. But mostly, I think any hype or speculation surrounding the results should be ignored, because the actual product of such an experiment will be much more about showing these limitations than it will be about creating an artificial soul.
Posted by: The Green Man at June 9, 2005 10:23 PMIt should be noted here that the researchers doing this have, to my knowledge, made no claims about creating an electric soul.
Posted by: flamingbanjo at June 10, 2005 07:23 AM