The Eureka Reporter

SUPERCOMPUTERS DUPLICATE HALF A MOUSE BRAIN

How close is a computer that can really think?
By Robert Reed
May 10, 2006

I really should be more concerned with local issues here in Humboldt County, but a BBC article published on April 27 was too interesting to ignore.  The title of the article is "Mouse brain simulated on computer".  A quick search on the web will bring it up.  A team of three researchers from the IBM Almade Research Lab and the University of Nevada have managed to create a simulation of "half a mouse brain" using the BlueGene L supercomputer.

According the the BBC, they ran a "cortical simulator" on the supercomputer for only 10 seconds because of the complexity of the simulation.  The BlueGene L supercomputer has 4096 processors with 256 megabytes of ram memory on each processor.  It's like having 2048 dual core computers all working together, with each one churning away the computations and an unimaginable rate of speed.  

Here's what caught my attention.  The contemporary modern supercomputer was only able to duplicate half a mouse brain at one tenth of the speed of real life.  

Some earlier similar research observed characteristics of thought patterns that were observed in some real mice.  Achieving this new level of being able to simulate half a mouse brain gives the researchers some new insights into how the brain works.  It's nothing to sneeze at.  However, I stopped reading the article and allowed my warped thought patterns to overtake my conscious thoughts on this peaceful Sunday morning of news browsing.  

Because I had indulged in one extra cup of high caffeine coffee, I started thinking about robots of the future.  I was thinking that we have not come as far as I had thought we were when it comes to research in making computers think like humans.  Long ago I lost my fear of a future of computers and robots that could think like people.  When that happens, computers will start making dumb colossal mistakes just like people.

It's like too much computing power is a bad thing. There must be a point at which artificial intelligence reaches a peak of perfection, and as it's computing power and thought processes grow, it will go "over the hump" and begin a long slow spiral of digression into human error-like failures.  If it thinks like a human, it has to screw up like a human, right?

How many smart people have you seen make unimaginable errors in judgment?  How about those really smart people who were at the wheel of Enron?  They were supposed to be smart with powerful brains.  And they convinced others with powerful brains to follow along, others such as the power brokers of Wall Street and countless investors, all of whom are supposed to be really smart people. 

When computers can feel greed, lust and power, look out.  Half a mouse brain is sounding better all the time.


There's another side to the question of how much is enough. Is half a mouse brain more than we humans now have? Some simple observations would suggest that this is definately the case. 

A little more coffee brought a logical conclusion to all of this. 

Maybe if we just built some robots that had half a mouse brain, it wouldn't be so bad.  I can think of a some people with brains like that.  Quite a few as a matter of fact.  I think that many times I fail to achieve the intelligence level of half a mouse brain.  Mice can remember how to run a maze and where things are.  I can't find things that are right in front of my face half the time.


My next thought came in a flash. What about most of our politicians?  Judging from their actions and not their words, one would have to conclude that the majority of them have far less than half a mouse brain. Political party affiliation doesn't affect this apparent fact of life.  Just look at what most of them do.  

Maybe we aren't as far behind in building artificial intelligence after all.

When we reach a full whole mouse brain, we might have something worth replacing some of our political minds with. In fact the BBC article mentions that the researchers are "tuning the simulation to make it run faster and to make it more like a real mouse brain."  This might be a goal worth attaining. And anyone with half a brain would know when to put a stop to all of it.  What kind of brain?  Your guess is as good as mine.

(Robert Reed is a columnist for The Eureka Reporter. Views and opinions expressed in this column do not necessarily represent those of The Eureka Reporter, its management or staff.)

BACK to the INDEX Home Page
BACK to the Technician's Masterpiece
BACK to the Eureka Reporter Column