Tuesday, July 26, 2005

Two hot items found at KurzweilAI.net:





Computer scientists to copy brain of a mammal

"Most theoretical neuroscientists have tried to simulate the brain using artificial neural networks. 'That is not our goal,' says Henry Markram, the professor leading the project. Markram wants to simulate the brain at every level of detail, even going down to molecular and gene expression levels of processing. At EPFL's Brain and Mind Institute, every facet of the brain is being examined and modelled."

Japan Plans World's Fastest Computer

"Japan wants to develop a supercomputer that can operate at 10 petaflops, or 10 quadrillion calculations per second, which is 73 times faster than the Blue Gene, an official of the ministry said on condition of anonymity."

Better computers mean better simulations. And better simulations increase the odds that we're inhabiting one.

4 comments:

razorsmile said...

"this I tell you brooother, you can't have one without the ooother ..."

W.M. Bear said...

From the mammalian brain simulation story:

To simulate even a single column requires enormous computing power -- power that, until recently, simply hadn't been available. The system that has been installed at EPFL has a peak processing speed of 22.8 teraflops and will use 8,000 processors to replicate just one neo-cortical column. This part of the project is expected to take two to three years.

This fact suggests to me that they are on a completely wrong track. They're trying to simulate thought processes by simulating brain structure. The point I've harped on here before is that mind does not equal brain, hardware (or in this case wetware) is not software. I wish them every unsuccess.

Kyle said...

wm bear -

Life on earth began as a collection of chemicals in a particular proportion, stimulated by a peculiar collection of external conditions.

If we are to in any way duplicate the human brain, we must first start with simple models using what we have...hardware...computers that can make zillions of calculations per second, since we know that the brain does this.

If we achieve a simulation which can process data just as quickly as a human brain, and with just as many stimuli to process, we will have a workable model.

From there...once the hardware is complete...theorists will rightly set forth on how to give the hardware "sentience". Initially, this may lead to a new kind of programming, where irrationality and "will" are provided a "sandbox" in which to evolve.

It might even go to the actual ingredients in manufacturing...small "semi-sentient" components which combine to form the more complex organism, etc.

At any rate, you have to have a simple life form before you can have a human being. Likewise, we have to build simple models of complex systems before we can duplicate them. But there is little doubt that we will eventually duplicate them, IMO.

If there is sentient life elsewhere, it is likely not like our own. Therefore, there's likely more than one way to make a sentient being. A good working model might even force us to question the meaning of sentience altogether.

Kyle
UFOreflections.blogspot.com

W.M. Bear said...

Kyle -- I simply think they're going at it from the wrong end, that's all. Modern computers were not created by first building complicated circuits and THEN figuring out how to program them. The logic (Turing, cybernetics, information theory) was developed FIRST and THEN progressively more sophisticated circuits were designed to implement that logic. If all you had was a computer running Windows and no source code, could you duplicate Windows simply by copying the microcircuitry? This is, in essence, what the Blue Gene and associated projects are trying to do. Because the mind-brain misidentification fallacy is so rife within the current scientific community, I sometimes get extremely frustrated trying to explain why this type of approach won't work. But I'm willing to bet money it won't.