Now we're arming robots with extra-superhuman synapse response? GREAT!

Lasers are not efficient for very short-haul communications. To communicate from chip-to-chip you need to turn high-speed electrical signals in to light and then back again. This is inefficient and generates heat.

If you can communicate from the silicon to the laser diode with electrical signals, why not go a little further and send the signal all the way to a neighboring chip?

While high-speed electrical signally has its challenges, there are techniques that could take it further than today's limits (of around 10Gbps for a SERDES link).

I would give them less than a 50% chance of success!

Incidentally, this is the same team that worked on capacitative chip-to-chip communications. While interesting, that had a number of challenges and to the best of my knowledge has yet to find its way in to any real products.

@LittleJon: Personally I would put my faith in Suns expertise over yours to decide whether or the technology is feasible, which it looks like they already have.
Everyone on the internet seems to be an expert...

I don't think it's a foregone conclusion our future computers will need photon-based (versus electron-based) pipelines to achieve artificial intelligence. It seems possible to achieve intelligence within a computational device even if that computational device uses a very slow transmission process (if you count my brain as intelligent--highly debatable I admit).

@LittleJon: It's worth noting that lasers are more hardened against EMP than regular wires--something the military would be interested in. Further, large arrays of CPUs use large amounts of wire. So the distance thing may be an actual factor here.

This idea isn't new but it does sound like they're actually going to pull it off this time. Kinda cool.

@xanderbeedle: OOOOOH! SNAP!

My favorite part is the graphic. :)

yeah, that whole 'speed of light' thingy is sooo limiting

SO it can play Crysis at 100k frames/sec? Cool.

Wasn't Intel working on the same concept but inside the CPU chip? building the laser in the silicon itself?
What ever happen to that?....

I would think LED's and light pipes would do much better.

One of the REALLY nice things about light is you can have multiple paths crossing each other without interference. Chips can become much more complex in design using light.

@LittleJon: Well, this technology is basically what IBM did here [gizmodo.com] and according to them, it consumes less power, not more, and I'm inclined to believe them. And btw, the laser diodes would be embedded in the silicon itself, your idea that you can send signals "a little further" is completely wrong.

The primary issue with this technology will probably be routing on the board. Signal propagation in optical mediums is adversely affected by sharp turns. Since most modern system boards are multilayer, this will represent some challenges moving the signals around the board. Hopefully they can compensate in the short term by having fewer signal wires with higher signaling rates.

@xanderbeedle: We actually don't know one way or another_ Except that Sun was able to persuade the Gov't to fork over $44 Million_ That's all we actually know at this point_

It may or may not ever come to be_ It may not work - but may end up spawning many other technology directions_ Who knows_

@scarbrtj: I agree with your viewpoint. The easiest way to create AI is likely by building a better model of the real thing, instead of trying to emulate it with a completely different type of infrastructure.

I'd envision that before going further with optics, the future of computing will first make a trip down the biological/chemical route. When we can simulate neurons that grow and develop within an immersive environment, interact organically with each other and use analogue language, then by improving synaptic communications, via light, the potential for intelligence is almost unfathomable.

@xanderbeedle: Actually I am an expert! I have a Ph.D. in electronics and I actually know Ron Ho, the Sun engineer quoted in that article. We were discussing working together (when I was at my last company) on something for their supercomputer project.

p.s. I've also worked on high-speed chip-to-chip signaling.

In best Dr Evil impression:
"I want fricking [computers] with fricking laser beams on their [chips]!"

@LittleJon: and now you just hang out in gadget blogs like the rest of us. How impressive.

Lasers = Heat
= Low Bandwidth
= Transfer rates
= CRAP

This is not a new idea. Has been around for many years! I give them < 25% of success.

This is not some Pokémon battle where a Bob "I have no life and I am a dork" scientist retorts like Scooby-Do "Uhhho Let's make lasers and f**k with it" and passes it by some nerdy "I have no clue about technology or trends" CEO and gets the stupid budget approved. Guys in real jobs know what I am talking about.

Sun has officially lost it!

@ltcmurray: Of course! Engineers like gadgets too! We don't spend all our time hanging around in labs wearing lab coats, you know!

I'm no expert in these matters, but I'm thinking if they can put multiple cpus together on one silicon wafer (it's just wafer thin, sir" :^) then why not put in the memory arrays, the gpu with all its supporting elements, and the physics chip, (if that really helps at all) and if that's too much to fit well on one wafer with wiring in between, why not use diode lasers? (can they send multiple packets simultaneously on different wavelengths over the same coordinates?) This could potentially speed that pesky fsb up considerably, no? and L1-3 cache could be added easily and I'm pretty sure that does make a difference! perhaps this technology, if successful, will make crysis finally playable on the very highest settings with everything turned on :^)) for that alone it might be worth the effort, even though crysis isn't nearly as much fun as farcry was, story/game wise (imo).

cheers.

Calling all researchers! Patent 5,051,738 is going into the public domain. I've personally seen this work and it does not need the sophistication of lasers. Please mention me when one of you hits it big with this as your base of further research. BTW, I've got other great patent mining if anyone is interested.

you know...there are SO many far fetching projects like this that we hear about on gizmodo...one of these days, i hope at least a few of them actually get done. a computer 1000 times faster than the current computers we have, and at a lower price would be more than a little breakthrough. The thing about it is though, how often does this kind of stuff happen? Breakthroughs are a seemingly rare thing. Sure they happen, but they tend to be pretty gradual, almost to the point where you expect them to figure out how to increase things like power and speed withing a given amount of time. Hopefully i'll be wrong though and within a few short year's we'll have super powerful computers that cost less than current ones do.

funny thing about this..i remember a computer teacher of mine telling the class about how this was going to happen one day...didnt seem too insane then, doesnt seem too insane now.

as for littlejohns comment..not saying hes the numbre uno engineer or anything, but i certainly think he seems to know his stuff.

@shawn_dude: More resistant to EMPS? Oh crap, we're doomed...

~Y

I never get bored with seeing "Pew Pew" on anything :]

It's like a breadcrumb to childhood.

please, what is the pr0n application for this?