IBM is prepped to lead the way into the next era of exascale computing, at least if the technology they showed off at a convention today in Chiba, Japan can live up to expectations. Today IBM lifted the veil on its CMOS Integrated Silicon Nanophotonics (CISN) technology at Semicon Japan, saying its next-gen silicon chips that communicate via pulses of light, rather than electrical signals, will be commercially available starting next year.
In the chip sector, nanophotonics are thought to be the inevitable future of CMOS chip tech, and nearly every major player in the field – Intel, Hewlett-Packard, Samsung, Freescale, etc. – has dabbled in the technology. CISN promises to unlock exascale computing power, removing the roadblocks between today's chip technology and processors that are 1,000 times faster than today's petascale computers.
How does one create a chip that executes one million trillion operations per second? Switch from copper to light. Chips that communicate over optical interconnects rather than electron swapping connections can perform much faster. And the more optical connections you have, the better. That's why IBM is currently expanding its manufacturing capacity (some of which is licensed to other companies) in order to produce the nanophotonic chips in large quantities. The company predicts CISN technology will ship next year, but in the next five years it's due for an evolution: It'll go from simply connecting systems to connecting boards in the same system and eventually to connecting cores on the same microprocessor.
How do these chips actually work? Are they still transistor based? What part of the chip is actually photonic? Is it just the communication between transistors, or between memory, or cores? Details!
another step towards the singularity
99.9999999999999% of an atom is empty space
In the second line, the Ninth word is, "Chiba." I do believe that you mean, "China."
No aid456 they really mean Chiba, Japan
Well if desktop PC's used nanophotonics does this mean just about anyone can decrypt classified/secret/private documents with so much computing power? Seems the age of spying may be coming of age!
Didnt Timothy Zahn predict this kind of computing?
I'll bet since these are all-optical chips, they dissipate little to no heat-so there will be no need for massive fans or liquid cooling setups. 10 years from now, photonic chips will make silicon chips look positively ancient.
I wonder if quantum computing will finally become mainstream by then also. Things look amazing for this field in the future. What will we do with computers with that much power?
(mulls to himself out loud)
I'm looking for Sarah Connor...!!!
So now we can snatch the fastest computer crown right off of China's head, right? That'll teach 'em! Go US computer tech.
@Voodoox I think you are totally wrong. I believe it is 99.9999999999998% space.
@aircraft_247 I am not sure who theorized but the idea has been around for a very very long time. decades.
@Golimaaar light generates MASSIVE amounts of heat. how do you think lazes shoot down missiles. let not forget this tiny thing called a SUN!!!! it does a rather mundane thing. HEAT the earth. i depends how the light interacts. and you still need energy to create light, and heat is energy. but you are right it will probably be more efficient in terms of wasted energy (ie: heat)
They will be one of the following:
Benevolent, in which case we are destined to be happy.
Malevolent, in which case we are doomed to extinction.
Indifferent, in which case we'll be ants and get stepped on from time to time when we get in the way.
But they are coming.
I don't think this is gonna mean anything to the home PC crowd for many years. As the article states they're still several years away from photonic inter-core communication and you're still talking about transistor/electron based cores. IBM has not built a photonic cpu but some kind of photonic transceiver to provide optical communication between systems, so yawn, roll-over and go back to sleep. That's what I'm doing.
A super-powerful cpu will no more become sentient than a super-monstrous combustion engine. Super-powerful computers and robots will kill us by pure random accident, but mostly due to being infected by malicious viruses created by humans.
A cpu is structurally designed to process information a certain way. Our human consciousness and emotion is a result of the structure of our brain circuitry and it's complex connectivity with sensory inputs and output controls. What I mean is that a processor can't become "intelligent" simply as a result of phenomenal speed.
You must construct additional pylons!
A super smart CPU - meaningless. CPUs stopped pushing the bounds of wha can be done some time ago. The move in personal electronics is to "less powerful than your CPU, but still more than you need for what you do - and you can carry it around." Here is what you will see -
1) Scholarly / industrial use (for jobs in climate, astronomy, genetics, etc that require massive computing power - and time)
2) Structurally (faster server farms for faster internet processing)
3) Reductionally (inproved computation on portable devices - catching them up to the usually overpowered CPUs that people seldom use a fifth of).
4) Gaming / entertainment (eventually the gaming market will increase game requirements to the rate of the next to top level of current limits - thus, the software is better, and drives the purchase of hardware - which is good for the software, as only those affulent enough to continue to adopt new systems generate real gaming profitability.
So, fancy processing, faster internet, a smarter IPocketEmptier, and an XBox 720.
Bow to your robot masters!!!
... read carefully.
They're still CMOS chips. They're supposed to communicate with *each other* optically. In other words, an optical bus between silicon chips. (As the article says in the last line, and also in noting that optics are replacing the *copper* pathways, not the silicon ones.)
Also, for those of you worried about The Future of Humanity ... shut up. = ) Sucks for modders, though - we're not so much moving toward the "singularity" as toward the printed computer, and I give us ten years on the outside until the last time you can pop open a computer case and see anything green inside. Far more likely to see a sealed aluminum wafer with a few data ports.
"What will we do with computers with that much power?"
Watch 3D porn.
Can I play Crysis on Max settings now?
Nanophotonic Chips are a good way to make processors faster and to market processing power. However, this will not make processors more intelligent. It will just make dumb processors dumber because of the chip’s incredible speed. I know this because I am a programmer. New programs need to be developed to run on the faster processor with Nanophotonic Chips. Ironically, there is no such thing like an AI (Artificial Intelligence) because the Intelligence can’t be artificial, period. My dog is a million, zillion, and google times more intelligent than all the processors running Nanophotonic Chips out there in any time. Intelligence is a metaphysical state where one can reflect on oneself and one can learn from one’s own experience. This requires a substance we call the spirit. How can the new and faster processor produce human spirit? How can the processor be self-reflective and organically learn from its own experience? How can it jump into higher state of being? If you are true evolutionary atheist, you may not get into faster processors because you will be confronted with the reality that no matter how fast the processor may be, it cannot evolve into anything without God’s intervention. On the contrary, the interesting thing is that the people who are the most enthusiastic about the potential of Nanophotonic Chips morphing into some kind of AI organism is the same group that adheres religiously to the notion of atheistic evolution.
Why should sentience require a "soul". You assume we have one. Your "soul" and "spirit" are products of your brain structure and chemistry. Your perceptions of metaphysics are artifacts of an evolutionary path that brought us to have culture, to wonder about the world and, having not yet found a way to explain it, to find metaphysical causes for what we observed.
No god or gods were needed for that, nor are any needed for AI. But I'm glad you warned us who believe that evolution requires no bearded men in the clouds that we will be confronted by what you have determined to be what is.
Now. About being a programmer...
Code monkey? Developer? Engineer? Researcher?
I guess all that academic, military and commercial research into AI is worthless because you say it can't exist.
I think you should tell someone before they waste more time.
We've looked very carefully at the brain and all we've ever seen is computation. Human brains don't have a 'soul' or 'spirit'; whatever the brain can do, a turing machine can do.
You may deeply resent it, but if p-zombies can exist, you are one.
we all know that the "Top Secret" branches of the military have probably been using this kind of tech for some time now. It's not like the gov't is decades ahead of the civillians by decades or anything like that..... Just another product of the reverse engineering (ripping off) alien tech from a51 or s7 or something like that.
Actually, "Chiba, Japan" is correct. China is a different country from Japan. Chiba is a prefecture bordering Tokyo, Japan.
Do you know what quantum computing is? Because this isn't it, therefore this technology has no bearing on whether or not quantum computing will be mainstream when nanophotonic computing is. This kind of computing is based around light being used to transfer data, though storage will still be pretty similar to what we currently have. Quantum computing is based on the theoretical ability for quantum events to be able to process and store information for data computing. It will actually be incredibly fast even compared to nanophotonic computing since it will happen on a much smaller scale and it will deal with quantum phenomena, many of which have been observed to appear and reappear randomly with no evidence of travel. The hard part right now is controlling the events and making them stable enough to process whole sets of data. Of course, not being stable enough to even do that means we are a ways off from being able to store data in that manner as well.
Definition of intelligence (per Merriam-Webster Dictionary):
a(1): the ability to learn or understand or to deal with new or trying situations.
By that definition, we have already created artificial intelligence in the form of agent programming. Not only have we created it, but it isn't a new creation. Agent programming was popular in the '90s. Agent programming allowed the device using it to learn from mistakes. For instance, take a robotic arm. If it tried to pick up an object, and it failed, it would examine why it failed and try again or realize it is incapable of picking up the object and it will no longer "waste it's time" with that type of object. The only reason agent programming isn't as popular as it used to be is because the code for it is so massive that any substantial learning requires an enormous amount of processing power. Since programmers didn't have the physical means to make it work in any complex situation, they started working on other ways to create the ellusive AI capability that agent programming promised to provide, but was too inefficient to be practical. This nanophotonic computing could very well provide that boost in processing power that is needed to use agent programming. However, with these new UAVs that are capable of piloting themselves and adapting to unexpected weather and other external events, I'm guessing a solution has already been found that works within our current tech capabilities. I've been out of programming for 3 or 4 years, so I'm not sure what's been done since them.
The intelligence you are talking about is actually the ellusive concept of the soul. The soul has not been proven to exist by the way, at least not as anything more than a combination of memories and self-awareness. Scientists used to believe that no animals were self-aware, but that has been disproven. Who is to say that self-awareness for a computer will not be possible with this nanophotonic technology, or the quantum computing technology that is sure to follow this new wave.
People crack me up thinking that the faster processors are the more likely they will become sentient. All being faster means is that it processes 1s and 0s at a greater rate. Artificial sentience can not occur while an AI is bound to the answer of 1 or 0. Quantum computing will likely yield true AI not so much as a result of its speed but its ability to allow a computer to answer a question with something other than 1(true) or 0(false). Quantum computing (grouped with a complimentary software advancement as well) will give the computer the ability to take shortcuts by predicting and guessing multiple outcomes at once beyond true or false (god help us all, he he).