March 08, 2008
Out of the Blue
Can a thinking, remembering, decision-making, biologically accurate brain be built from a supercomputer?
Jonah Lehrer in Seed Magazine:
In the basement of a university in Lausanne, Switzerland sit four black boxes, each about the size of a refrigerator, and filled with 2,000 IBM microchips stacked in repeating rows. Together they form the processing core of a machine that can handle 22.8 trillion operations per second. It contains no moving parts and is eerily silent. When the computer is turned on, the only thing you can hear is the continuous sigh of the massive air conditioner. This is Blue Brain.
The name of the supercomputer is literal: Each of its microchips has been programmed to act just like a real neuron in a real brain. The behavior of the computer replicates, with shocking precision, the cellular events unfolding inside a mind. "This is the first model of the brain that has been built from the bottom-up," says Henry Markram, a neuroscientist at Ecole Polytechnique Fédérale de Lausanne (EPFL) and the director of the Blue Brain project. "There are lots of models out there, but this is the only one that is totally biologically accurate. We began with the most basic facts about the brain and just worked from there."
More here.
Posted by Abbas Raza at 12:14 PM | Permalink
Comments
Abbas, unfortunately when I click on the more button IE tells me the page can not be found. I would like to read more however so I'll check back later to see if it reappears.
Posted by: rhbee | Mar 8, 2008 4:06:37 PM
the correct link:
http://www.seedmagazine.com/news/2008/03/out_of_the_blue.php
Posted by: a | Mar 8, 2008 5:27:25 PM
Thanks rhbee and a. I have fixed the problem.
Posted by: Abbas Raza | Mar 8, 2008 6:10:30 PM
Forgive a silly philosopher for poking my head into places I obviously don't belong, but suppose a human brain is modeled in ten years and successfully placed inside a robotic body.
How, exactly, do we verify that the thing we've created is having experiences? What, do we just ask it?
"Robot! Are you having experiences?"
"Yes, of course I am. What a stupid question."
The programmers seem unbelievably cavalier in insisting that such a model will be sentient, but they offer nothing but platitudes ("the mind is just a lot of information processing!") to support this.
Posted by: Nick Smyth | Mar 8, 2008 8:29:20 PM
Right on, Nick.
Of course, what it means to be human is continually being defined downward, so that it gets easier and easier to be able to say we are capable of creating "artificial intelligence."
To what end? The result is that we believe ourselves less and less capable to determine what kind of culture we want to live in, and we lurch (slouch?) towards a putatively inevitable future that nobody wants.
Posted by: Chris Schoen | Mar 8, 2008 8:42:34 PM
We have not been capable of defining "consciousness" nor have we been able to determine the locus of the essence of what make us "sentient" beings.
This "brain new world" is like an automatic pilot: automatic, yes. Pilot? No.
But scientists in their hubris will continue to beat the dead horse of their omnipotent dreams to arrive at the "putatively inevitable future nobody wants" --- as Chris Schoen aptly summarizes.
Posted by: Felix E. F. Larocca MD | Mar 8, 2008 10:19:40 PM
We have not been capable of defining "consciousness" nor have we been able to determine the locus of the essence of what make us "sentient" beings.
This "brain new world" is like an automatic pilot: automatic, yes. Pilot? No.
But scientists in their hubris will continue to beat the dead horse of their omnipotent dreams to arrive at the "putatively inevitable future nobody wants" --- as Chris Schoen aptly summarizes.
Posted by: Felix E. F. Larocca MD | Mar 8, 2008 10:20:29 PM
So eventually, in 10 years or so, about the same time that Hill/Obama pulls the troops out of Iraq, the blue brain folks will have created something from the study of how many rat brain pans? Meanwhile, how many million books will have been written by real human brains that will also have to be understood during that time? I can see (ha!) inside a human brain every time I open up something to read or turn on the tv or go to a movie. It seems to me that this project is just trying to have a baby without the sex. Ugh! And they seem serious about not doing it just for the side effects of learning something about brain function along the way,too.
Abbas,
I enjoyed the article but I really don't see what the need is for the project.
Posted by: rhbee | Mar 9, 2008 12:39:42 AM
It's not "bottom-up." Nobody has seen the bottom--or more than a tiny patch of it. And if it's not necessary to see to the bottom, nobody knows how deep or in how much detail we need to know the physiology in order to explain mental phenomena. That's why such a project is worth doing largely--as a test of whether we are anywhere near to knowing enough detail.
Posted by: mt | Mar 9, 2008 4:30:05 AM
When Isaac Asimov first envisioned thinking computers in his sci-fi stories, he assumed that these would be so massive and expensive that only the huge, benign government agency in his rather simplistic future world would have the resources to create them, and so would be able to lay down ethical rules (his 3 Laws of Robotics) for them to think by. Today Supercomputers are becoming widely available and affordable, and if anyone ever does manage to make such a machine ‘think’, nobody will be able to enforce any ethical standards.
I rather agree with the previous commentators, that the goal of AI is essentially undesirable, but the race is on, whether one approves or not.
I have the feeling that the quest is going to prove somewhat simpler than it looks, and that someone will come up with a set of algorithms that enable a computer to build its own models of reality and begin programming itself. From that moment, I think there will be an explosively rapid development towards very smart machines. I base this prediction on the fact that the basic structures that make up human brains evolved hundreds of millions of years ago in brains that were much smaller, with many fewer neurons than are simulated in the supercomputer described here.
Forget about poetry and imagination: anyone who has observed how elderly people behave and move will have noticed just how predictable they are and how many of their actions and thoughts are simply recorded and replayed many times over. How long before robots are able to mimic your granny? How long before they are able to efficiently manage a supermarket checkout or clean a home?
Posted by: aguy109 | Mar 9, 2008 5:17:30 AM
Don't downplay the philosophies and thoughts behind many of the scientists working on AI, some of whom I personally know. These are extremely smart and well thought out people, no matter the selected quotes in this article. You may not be personally comfortable with a machine that can out-think you (I'm certainly not), but that doesn't mean others haven't come to terms with the idea. In fact, many are extremely enthusiastic.
Nick, what is sentience? How is it different from intelligence? You seem to think it has something to do with having experiences. I doubt you would deny that Animals experience the world around them. Yet defining the level of organism at which experience evaporates becomes difficult when you start to progress down the tree of life. Does a rat have experience? Yes. A Fish? They certainly don't like it when I hook them. A tadpole?
The argument can be made that what makes humans and some animals particular is their ability to reflect on, and control, experience: consciousness. But here again things quickly become muddled. There have been many interesting experiments that show the cerebral cortex lighting up after the motor cortex has given a command, with the ability of different individuals to countermand that command differing. If we can't always control our own actions, do we act in a conscious manner?
Maybe consciousness is the ability to analyze non-immediate situations. This too is a problematic statement, at least from an anthropomorphic view, as computers already excel at this task. A project like blue brain will put them over the top.
Very little is clear to me with respect to computer intelligence. I think many of the claims of general AI coming from simulated brains, brains we ourselves can not yet explain with much detail, are overblown, at least in terms of time-frame. But society needs to start thinking about these issues now, and those smarter than I need to start framing them in a more intelligent manner. Humanity will have to deal with "artificial" intelligence at some point, and that point is sooner than later.
Chris-
Get over humanity. We're but a blip on the galactic scene, soon to die out in the grand scheme. This doesn't make me value my or other's lives any less, but it's certainly more realistic than expecting humans are, in some way, a special species, imbued with something magical that sets us apart. Understanding that makes us a no less capable species, just a more reasonable one.
Posted by: Cyrus Hall | Mar 9, 2008 9:00:42 AM
aguy109,
Funny you should mention Azimov because everyone thinks of his I, Robot right away when the subject is AI. But David Gerrold's, When Harlie Was One, is also worth reading if you want to speculate and certainly, Clifford Simak's City will take you somewhere too. These days though I am enjoying the trials and tribulations on the RU Serious space craft and Sarah Connor's attempts to corral the computers on her Chronicles. Do you suppose that the blue brain people read these too?
Posted by: rhbee | Mar 9, 2008 12:01:13 PM
Chris,
I, uh, didn't say anything about animals. Their consciousness/subjective experience is equally mysterious. All experience is mysterious. That's my point. I'm not singling out human experience as special.
Forgive me, I'm having a hard time seeing what your Carl Sagan-ish view of humanity as a "blip" has to do with my point. The "hard problem" of consciousness is almost universally recognized as THE most difficult problem of modern philosophy. To dismiss it by just assuming that experience equals intelligence is absurd.
Posted by: Nick Smyth | Mar 9, 2008 3:03:17 PM
Nick,
Are you confusing Cyrus and me?
Cyrus,
I never said humans were special in regard to other life. As far as I know building an artificial cockroach may be as impossible as simulating human sentience.
As for your argument that others are "comfortable" with the prospect of AI, so what? We don't live in an intellectocracy, or at least not explicitly. At any rate I'm not alarmed by the prospect (though I think it's a waste of money and a waste of rats), just skeptical. Nick has already raised a number of challenges to the Blue Brain claims; I'll add one more: What sort of "experiences" can a brain without a body, without sense organs, without hormones, possibly have? Markram wants to put a simulated rat brain into a robotic rat body, but how does he propose to simulate both the motility of an organic rat, and the incredibly sensitive facility for sensory input?
On top of that, how will the rat generate an interpretive apparatus? A living rat acquires instincts genetically, and other behaviors through socialization. Processing is just the beginning of a sensory experience; next is the matter of associating data with memory, instinct, desire.
We can't stop our scientists from pursuing these projects if this is what interests them, but we may be able to educate future scientists better about how to spend their time, given the challenges that face our biosphere.
Posted by: Chris Schoen | Mar 9, 2008 3:58:31 PM
Oops. Sorry Chris, yeah, that was aimed at Cyrus.
Your point about sensory input, though, is a good one. Having just come out of a seminar on the visual system, I can say with all confidence that no-one has any friggin' idea what sense-data is, empirically or philosophically. Until THAT gets sorted out, a giant collection of neurons is not going to model a living organism of any type.
There might be dualistic assumptions running through this line of reasearch, insofar as the importance of EMBODIMENT is neglected, as you point out it is.
Cyrus: it also occurs to me that your "get over humanity" invective might also be aimed at the Blue Brain team: why the incessant drive to model a HUMAN brain? If we're so confident that information processing is all there is to consciousness, why not just build any old colossal supercomputer and let it run?
Posted by: Nick Smyth | Mar 9, 2008 6:43:43 PM
I will respond in more detail tomorrow when I have the time, but I'm afraid my main arguments were somewhat misunderstood. For now, let me readdress my last point about the non-special place of humanity, and the twist Nick has thrown on it.
Indeed, Nick's point is excellent, and is one of my consistent criticism of the current work being done. There are many significant architectural and procedural problems ("mistakes" if you will) in the human brain, and it seems a little silly to copy them directly into computer AI. The problem, of course, is that many of the mistakes the brain makes are not fully understood, and AI researchers don't want to wait. And so the current efforts have an edge to them which is rather ominous. Do we really want super-intelligences far smarter than us, with all the same flaws, that we don't completely understand from a theoretical point of view? I think not.
I'll try again on the other points tomorrow.
Posted by: Cyrus Hall | Mar 9, 2008 8:29:27 PM
Of course, what it means to be human is continually being defined downward, so that it gets easier and easier to be able to say we are capable of creating "artificial intelligence."
I want to say it's more like the opposite. Time was when a computer needed to recognize speech/faces, play chess, perform complex calculations or navigate unaided to possess some AI. Every one of those barriers has eroded, and each time we've defined intelligence upward. AI researchers only half-jokingly define intelligence as that which silicon cannot yet do.
Posted by: D | Mar 10, 2008 7:53:56 AM
This sentence in the third paragraph caught my eye:
"The behavior of the computer replicates, with shocking precision, the cellular events unfolding inside a mind."
I'm waiting until the computer replicates the cellular events unfolding inside a mind "with shocking im-precision" for it to get really interesting.
Posted by: doug l | Mar 10, 2008 6:59:07 PM
Post a comment