IBM in "we're going to build a computer" shock

[U-96]

FIVE YEARS to build
ONE MILLION processors
ONE PETAFLOP!

Someone build an RC5 client please?
http://news.bbc.co.uk/hi/english/sci/tech/newsid_553000/553690.stm

[welsh wizard]

U-96
By the time IBM complete it, it will be out of date, (what happened to that chip company they owned?) http://www.sysopt.com/forum/smile.gif
WW

[U-96]

Haha, WW, how true... as I was writing my original post, I was chuckling to myself thinking, "I guess single AMD processors will prolly be like that in 5 years" http://www.sysopt.com/forum/wink.gif
Or it could gain sentience it 5 hours, control Skynet, and nuke the planet.

All it needs is one major scientific breakthrough to destroy a project of that length.

[Dominus]

****, that makes my Athlon 666 look like a second rate Commodore.

One other point to make on the SkyNet issue: Computers cannot evolve unless they can die, or their programs have some sort of death equvalent, like genetic programming. But if the gov't WAS stupid enough to run GP on that machine, it'd still probably take a year or two for an AI to evolve to the capacity of a chimp. But it's a step in the right (wrong?) direction at least.

[welsh wizard]

I wonder what would happen if we got all the worlds old Commadors linked together if we could put IBM to shame http://www.sysopt.com/forum/wink.gif
WW

[CMonster]

I should tread lightly on this sacred cow.

The only thing that becoming "sentient" or supposedly being able to "evolve" in terms of AI would prove in a computer of this magnatude is that a huge amount of planning and purposeful design went into the "creation" of it in the first place - thereby making somewhat mute the idea that it could evolve in the first place.

And the idea that organic life is any less complicated also seems somewhat put to rest by the idea that it takes such enormous computing power to anaylize even "simple" protiens, let alone the mysteries of some number of billions of precisely sequenced chemical bonds in a DNA strand.

As for me it is back to the slime......

[This message has been edited by CMonster (edited 12-07-1999).]

[welsh wizard]

Her's a good one for you , a fellow trader I know came for a visit and we were talking about this IBM, he reckons he's seen a site out there that talked about some computer that was protein based, we joked about whether it needed prime steak or veggy matter to keep it going, he promised if he can find the site again he'll e-mail me the http: address, if he does I'll post it.
WW

[BluMoose]

Computers won't have the power comparable to that of the brain for probably another 40 years or more (assuming Moore's law, that computing power doubles every 18 months holds). Since AI engineers consistently push the limits of current technology I wouldn't be surprised if human-like AI came out around that time (and products like automatic vacuum cleaners/ect that don't need that power will come before). The design would be similar to the human brain in that it would be modular. That is different parts would have different functions, such as analysis of visual input and walking (with sensory feedback) routines.
As for computers analyzing proteins, that refers to long proteins and how the fold. Protein folding is the resultent effect of the atoms in the protein and the forces they exert, particularly their electorns. This can get quite complicated because the general path electrons follow can change dramatically when an atom forms a bond with another atom. However, this isn't relevent to the issue since current projects suggests that you can mimic the effect of a biological system without mimicing the process that it uses.
As for a protien computer, I am not aware of any such device, though there is research going into DNA computers. The premise behind this type of processing is that DNA strands when placed in a bathe of the components that make of DNA would compete, and the resultant DNA of the greatest frequency would be the optimum answer to a problem. This has been successfully used with the travelling salesmen problem (a salesmen has to travel to X different cities at varying distances from each other, the problem is to figure out the shortest path that goes to all cities and there is no known mathematical formula/procedure to figure it out).
You can find some information on this in the December issue of Scientific American, as far as AI goes, I would have to look into protien folding and DNA computers. Though I think both have also been in Scientific American.

-BluMoose

[This message has been edited by BluMoose (edited 12-07-1999).]

[RobRich]

Here's some interesting links about computers and AI:

laws.lanl.gov:80/XCM/neural/neural_announcements.html (http://laws.lanl.gov:80/XCM/neural/neural_announcements.html)

PLus if you think AI and bio-mechanical computers are cool, you need to check out quantum based computing:

qso.lanl.gov:80/qc/ (http://qso.lanl.gov:80/qc/)

[U-96]

Biomolecular computers: http://news2.thls.bbc.co.uk/hi/english/sci/tech/newsid%5F370000/370035.stm

Quantum dot power http://news2.thls.bbc.co.uk/hi/english/sci/tech/newsid%5F314000/314503.stm

A lot of this month's (Dec 1999) Scientific American is worth reading, particularly the stuff on AI.
Humans are good at stuff like recognition rather than maths because we're hardwired that way from eons of evolution seems to be the main thrust of the argument.
To see how cool humans are: http://news2.thls.bbc.co.uk/hi/english/sci/tech/newsid%5F305000/305411.stm

U-96

[welsh wizard]

BluMoose
when you started talking about dif cpu's for dif parts of body, brought back memories of the old Atari 8 bits with all the dif cpu's that were inside it to hadle dif things, the one I remember best was the Antic chip a cpu in it's own right running with the main CPU. http://www.sysopt.com/forum/wink.gif so have we progressed that far yet.
Pentium lll keeps getting more and more added compared to the very first Pentium classic only a few years ago. http://www.sysopt.com/forum/smile.gif
WW

[BluMoose]

Too bad those BBC articles don't always give that much information (especially the biocomputer one, though it does sound very similar to nanotechnology).
As for "dif cpu's for dif parts of body," that is really how our body does it; our brain is made up of different modules which process different bits of information. It is also the most practical, since computers would need to be far faster and complex than the human brain if you only had a general purpose CPU, making specialty sections allows those parts to do their job better (like a 3D graphics card in a computer).

-BluMoose

[Apostle 83]

Wow, definately some interesting stuff here. I doubt though, that quantum comps will be cheap enough

[Apostle 83]

Wow, definately some interesting stuff here. I doubt though, that quantum comps will be cheap enough to ever use in home comps. Organic systems are making headway, they just have to fully understand HOW WE WORK! Maybe a year and a half ago (max), I remember reading that they had mapped out the entire neural network of earthworms and sea slugs. Its just a matter of time- once processing power begins to truly splurge, it'll snowball and be REALLY, REALLY COOL! http://www.sysopt.com/forum/smile.gif

[U-96]

I agree BluMoose, the BBC articles aren't hardcore science, but they usually give good links to the original sources.

The article I actually went looking for (but couldn't find) was a little more relevant and concerned an experiment (at MIT I think) about a couple of scientists who had made a true organic computer, consisting of a few nerve cells. I think it worked like this:

1. One electric stimulus into cell #1
2. Two electric stimuli into cell #2
3. Cell #3 tranmits three times

Yep, the cell adds up!
From small acorns grow great oak trees...

U-96

[Banti]

Dominus,
In regards to the Skynet thing. Computers do not have to die. They could behave like the nanites from Star Trek TNG. They reused parts from older units to build new ones in order to 'evolve'