Item — HN Reader

Important subject, but this article sucks.

Instead of just two states, on or off, as with transistors, memristors can represent many states. This means we can create new types of computing models, we can also create analog computers, which you don’t program, but you let them learn. You can then replicate the learning to other memristor analog computers.

- We might be able to use memristors in a similar way to synapses in the human brain.

_snort_

2 comments

For more technical details try this video of a keynote by R. Stanley Williams: http://www.youtube.com/watch?v=bKGhvKyjgLY

Submitted as a YC story here: http://news.ycombinator.com/item?id=1287576

What is wrong with the article?

This bit is rather silly:

"This means we can create new types of computing models, we can also create analog computers, which you don’t program, but you let them learn."

Analog computation is nothing new and has pretty much been superseded by stored program digital computers - for very good reasons.

I used to knew am engineer who had used a hydraulic analog computer to solve systems of differential equations that modeled fluid flow... :-)

I'm not claiming to be knowledgeable, but isn't it true, that analog computers are easy and fast to use for computations (sliding logarithmic ruler, self regulating hydraulic systems,...). But notoriously hard to construct IF you're looking for something multipurpose and programmable.

It's mostly the "you don’t program, but you let them learn" part that's a bit weird/inaccurate. You have to program analog computers, and don't get any automatic learning built in. It's just a different kind of programming. Some things, from a digital-programming perspective, are amazing freebies that would require difficult programming on a digital computer. Other things, that on a digital computer would be trivial, are frustratingly difficult to implement. But it's still programming of a sort either way.

And it doesn't make a ton of difference these days anyway, because "continuous" vs. "discrete, but with trillions of gradations" are indistinguishable for a lot of problems, so you can simulate an analog computer on a digital one, or vice versa, with reasonable accuracy.