2001, A SPACE ODESSY

[Jeanway 0]

Is that a reality now? A computer that can take over everything? I love that movie.

[WEAREBORG4102 0]

Not even close to rivaling the human Brain...

According to Moore's law, computing speed doubles every 18 months. right now a super computer can perform about 10 billion calculations per second. That's a super computer.... they do not apply to moore's law. One single cell in the human body can caluculate around 4 quadrillion caluculations per second. One cell!!!! Imagine your brain!!!

[Lubak10 0]

Not even close to rivaling the human Brain...

According to Moore's law, computing speed doubles every 18 months. right now a super computer can perform about 10 billion calculations per second. That's a super computer.... they do not apply to moore's law. One single cell in the human body can caluculate around 4 quadrillion caluculations per second. One cell!!!! Imagine your brain!!!

woooooooooow! That's kinda freaky! I don't think mine's worken that fast, though. Maybe I've been doing to many times. If so then I blame my sister.

[Jeanway 0]

Yea, I kinda thought that too. With all the systems contolled by the brain plus it takes care of itself, it's pretty amazing. I don't think we will ever have a computer comperable. But do you think a computer can learn, on it's own? I mean, uh, help me out here someone. Hmmm, perhaps I not saying this correctly. By a sereis of random programs it may eventually do something, not bad , but unexpected? Cross-referencing and........Understand?

[nik 0]

What you're describing Jeanway sounds a lot like what the computer world calls "genetic algorithms." We use them a little bit where I work. Some of the

experiments we do can be pretty elaborate and in some cases very difficult to compile the data into useful structures. For that, we use a program which essentially takes "program fragments" and compiles them into an analysis package. It keeps doing that until it gets something useful (based on minimization routines). Is it intelligent? Yes. Compared to humans? No. In this case, we have rudimentary (on a biological scale) pattern recognition routines, but bad compared to those of humans, and we trade speed for that. One drawback is that the

algorithm may never "converge" on a solution, giving erratic results. (We always joke about a Nobel Prize experiment being botched because the software keeps giving us video games.)

Other applications that I've used are neural networks. Back in the days when we

were strugging to find the top-quark, we had huge amounts of data (actually small

by today's standards), and we needed a fast way to filter through this data and characterize particle decay jets, rank them based on their appropriateness,

eliminate the bad ones, and suggest good ones. The neural network, which is

a multi-level weighted parameter feedback algorithm, is what we used. We

used a "hardwired" network - that is, we burned a chip about four times the size of your Pentium IV rather than process data in software. The result is that we

effectively had a dedicated "program" operating about 10-100 times the speed of a

normal computer.

Today, we still use neural networks, and hardwired genetic algorithsm, but we also

use things like Application Specific Integrated Circuits (ASICs) and Field Programmable Gate Arrays (FPGAs). All this stuff is really just tools to us

scientists, but we have to virtually have engineering educations as well as our

education to be scientists.

[nik 0]

...do you think a computer can learn, on it's own? I mean, uh, help me out here someone. Hmmm, perhaps I not saying this correctly.  By a sereis of random programs it may eventually do something, not bad , but unexpected? Cross-referencing and........Understand?

By the way, this is an impressive use of intuition, Jeanway. This is sort of thinking has been known to spawn great ideas.