Tiny Brain-Like Computer Created

[D3B]

This sounds neat.

The most powerful computer known is the brain, and now scientists have designed a machine just a few molecules large that mimics how the brain works.

So far the device can simultaneously carry out 16 times more operations than a normal computer transistor. Researchers suggest the invention might eventually prove able to perform roughly 1,000 times more operations than a transistor.

http://us.rd.yahoo.com/dailynews/li...nce.com/technology/080310-brain-machines.html

 

[Npl]

so, if it mimics the brain, does that mean its gonna have the same amount of errorenous operations?
Could it be that in the future you buy a AI-Lawnmower that has legasthenie and move towards the neighbours lawn instead of yours?

Im just saying that directly comparing brain(-like) to the processors we have now aint possible. A small Pocketcalculator beats my calculating skills by far, not to mention Software like Maple, even if I tell myself im smarter than the dumb machines dozends of times.

 

[pcchen]

From the article, it looks like "mimic brains" means only about the ability to easily transmit the result of a "transistor" to multiple transistors. Current circuits do not like large fan-outs because it can reduce performance.

However, to actually work like a brain, the connections will have to be dynamically created/destroyed. I don't know if it's possible to do so with this special molecule. Fortunately we do have an age old and proved way to create new brains, although it takes time.

Of course, this is still interesting as currently transistors are already near a wall. We certainly need discoveries like these to pave roads for future transistors.

 

[3dilettante]

It's not clear to me that the new setup necessarily improves fanout issues more than if we were to magically shrink a 16-fanout transistor down to that size.

The test setup ties the molecule to an external needle, so there's an entire lab setup and essentially unlimited power source tied to this one puny little molecule.
I'm sure standard transistors would do fine with 16 fanouts if each was tied to its own VRM at some arbitrary voltage.

The 4-state molecule proably quarters noise resistance at the very least as well.