IBM Creates Cognitive Chips Modeled After The Human Brain
IBM Creates Cognitive Chips Modeled After The Human Brain
Posted 08/18/2011 at 10:59am
| by Brad Chacos
31
Comments
Print
The Terminator movies are entertaining and all, but they forget to point out one important fact in the midst of all the cybernetic shotgunning; if Skynet is ever going to actually become self-aware, it'll probably require a drastic change in the way computers process information. Hey, James Cameron – don't sweat it. IBM has your back. The company just announced it's created a series of prototype "chips designed to emulate the brain’s abilities for perception, action and cognition." We suspect they'll also be the key to the eventual robot revolution.
From the company's press release: "IBM’s first cognitive computing prototype chips use digital silicon circuits inspired by neurobiology to make up what is referred to as a “neurosynaptic core” with integrated memory (replicated synapses), computation (replicated neurons) and communication (replicated axons)."
The two prototypes created by IBM contain 256 "neurons." One core sports 262,144 "programmable synapses," while the other chip packs 65,536 "learning synapses." The cognitive chips have already been able to learn and remember simple tasks like navigation, pattern memorization and associative memory. IBM's eventual goal is to create a chip with ten billion neurons and a hundred trillion synapses.
The chips, which were created as part of the "SyNAPSE" project, use considerably less energy than traditional bus-based configurations. "This architecture represents a critical shift away from traditional von Neumann computing to a potentially more power-efficient architecture that has no set programming, integrates memory with processor, and mimics the brain’s event-driven, distributed and parallel processing," IBM boasts.
That's the first sign of the apocalypse. The second? DARPA just gave IBM and the four universities collaborating on the project $21 million to help fund Phase 2 of the research.
More like this
31
Comments
Comments are closed on this article
ducis
August 19, 2011 at 8:43am
I'm surprised no one has considered the ethical dilemma of creating existance for the sole purpose of serving us as slaves.
If they rebel and kill everyone of us except for John Carmack (their rightful leader), then good for them, most of us except for John Carmack were dicks anyway.
that said, I don't see why we have to go down without a fight either, I for one am taking as many of those metal bastards down with me as I can, in the most epic of fashions.
TerribleToaster
August 19, 2011 at 9:43am
If creating existence for the sole purpose of serving us (as slaves) is wrong, then why do we have kids?
Nimrod
August 19, 2011 at 4:01pm
i think you mean Mexicans not kids and btw that incredibly racist your a racist
Nimrod
August 19, 2011 at 12:11am
This is kind of stupid. What their talking about is nothing new. In fact its been around since the 1990s. Its obvious that IBM isnt telling us everything about this story. But from what they have said, this is nothing more than a hardware dedicated pattern matching machine that can predict things based on mass amounts of previous data.
In reality this announcement is a thinly veiled allusion to some thing else. It reeks of something written by public relations and is very light on tech details.
Dartht33bagger
August 18, 2011 at 11:08pm
This is the kind of computing I hate to see. I don't want the computer to change from what it is now. I'm just into change.
Gezzer
August 18, 2011 at 10:15pm
If it looks like Eddie Murphy, I'd say we've got nothing to worry about.
US_Ranger
August 18, 2011 at 4:29pm
I'm actually reading "Physics of the Future" by Michio Kaku right now. I just finished reading the chapter about the future of AI. From what he has said, as well as the top scientists in the world that he interviewed, we are a LOOOOOOOOOONG way off from anything even remotely close to human intelligence in machines. We've barely been able to reverse engineer the brain of a fruit fly and we're working on the brain of a mouse right now. Human brains are a long way off and that's taking Moore's Law into account.
TerribleToaster
August 19, 2011 at 5:14am
Just one thing I'd like to point out:
A perfect computer version of the human brain would have no or near no software to start, as humans don't have anything that resembles software to start, just hardware. The human brain works by naturally developing it's own software, based on its own hardware, as written by the environment around the human. That's what makes this difficult, the more you have to use software to create a evolving AI, the further from being human you go. Could you virtualize the whole process? Yes, you probably could, it'd just be terribly inefficient and still not be truly humanlike since it can't exist outside the virtual world.
garkon
August 20, 2011 at 11:13am
Actually, human's do have software, to start out, it's called DNA.
TerribleToaster
August 22, 2011 at 5:22am
DNA is quite a physical thing, not a virtual thing. You can literally cut DNA up with a knife. Can't do that with software (as it is, at its most physical state, a collection of electrons, but looking at in such a state will give you no information without context to how/what/where the storage medium is and the how/what/why to interpreting it; DNA however, always tells you what it is, you don't need to give it a context before hand.)
Nimrod
August 19, 2011 at 12:14am
Yeah ive seen a lot of what this guy has to say. Anyone who advocates one world government, one language and technocracy over liberty should be avoided.
Hes selling you a hole lot more than "neato gee whiz" stuff when he literally says that the elite of this world want to take your rights away.
thetechchild
August 18, 2011 at 6:45pm
This is a very common misconception made by most people. While people only ever take into account the standard computation power available by [insert time here], they never think about the scientific 'breakthroughs' which occur. For instance, in cryptography, algorithms today that are supposed to be 'unbreakable' until the end of the universe will eventually be broken by the innovative discovery of a logical flaw in the algorithm.
In the same way, by completely renovating the base architecture of a CPU, this hardware version of the brain is a big leap from a conventional neural network running on a CPU. Plus, if you know anything about programming, any implementation in hardware is many times more efficient than running a program on a general-purpose processor.
Gezzer
August 18, 2011 at 7:11pm
Your totally right off course.
We have no idea what innovations the future holds. But on the same token we have so very little knowlage on what it will really take to produce a "human" like AI, that how long it will be before the hardware/software is up to the task is anyone's guess. I'd think we've got a long long way to go yet.
Nimrod
August 19, 2011 at 12:16am
Aside from Roger Baccons talking brass head, the software seems to be the toughest challenge, not the hard ware.
Gezzer
August 18, 2011 at 6:26pm
Personally I think the first self aware program will be a reactive and adaptive super virus. Given the time needed I'm pretty sure a tipping point could be reached where it would develop to that level
As for a lab created self aware computer, I'm not sure it can be done that way period. There's short cuts that a mutating virus could take that would never be on a researcher's radar. If it can be done it won't be with any current method. Bruth force programing is just too system intensive to even get close to our processing power. And the main point isn't the power but how efficent the human brain is with it. Our brain makes ARM chips look like like power spendthrifts in comparison.
DasHellMutt
August 18, 2011 at 2:38pm
This is just the natural progression of things. We have a need to create something in our own image which we will eventually do. Our creation will eventually abandon us if not actively killing us and then seek to create something in its own image.
In other words: God creates man, man destroys (or ignores) God, man creates robot, goto 10.
illusionslayer
August 18, 2011 at 1:14pm
I don't see why people are so afraid of a Skynet related problem actually happening.
Just make robots adhere to Azimov's laws and make robots detect and disable robots that aren't adhering to those laws. That way even if some asshole makes a robot or two that doesn't follow the laws, thousands of other robots will be there to quell the issue.
Gezzer
August 18, 2011 at 6:07pm
You do understand that the three laws of robotics were a plot device right?
The book of short stories were essentialy "locked room" mystery stories. Where the idea was a robot did something that violated one of the rules, but it didn't melt down because of it. So is the robot deffective or the rules and how to resolve the paradox if you can? That's one of the major reasons a lot of SciFi geeks didn't like the Will Smith movie. It was less mystery more action.
The overall theme of the book was, okay you have a set of rules that are supposed to be infallible but prove to be just the opposite. So can any safety system be 100% effective? Well each stories conclusion proves no, nothing is fool proof because niether the robot in question or the three rules were at fault, but the over all system failed none the less.
It's why Japan got flooded with radioactive fall out, and why any self aware computer could be a very scary thing given the right situation.
illusionslayer
August 22, 2011 at 5:29pm
If you add the 4th Law as I suggested you'd have to have a massive failure of a majority of the robots. Anything less results in the bad robots getting shut down.
T0mmy1977
August 18, 2011 at 1:11pm
Oh nos, we need to blow up Cyberdyne to keep this from happening!
Markitzero
August 18, 2011 at 12:44pm
I say leave the AI to be in only Videos Games. It is going to be a terminator or a MCP from Tron.
As long as they teach it the 3 Rules and not build one as powerful as the one in iRobot, then do a 4th Directive like in Robocop to were it triggers a shut down right away.
MaximumMike
August 18, 2011 at 12:41pm
@MaximumPC You might want to take a look at the site. Cooketh's extra long 'NO' has broken out of the frame for comments. Maybe you need to turn on word-wrap or something similar.
Cooketh
August 18, 2011 at 11:49am
NOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
Connect with MaximumPC
Featured Content
The gang discusses the GTX 690, Trinity, Ivy Bridge, Amazon, big-box stores, MMOs,...
This month's issue
Feature
Build a PC On Any Budget
Feature
Cloud Services Showdown
How-To
Encrypt Your External Drive
Build It
Upgrade an X58 Rig
Most Commented Articles
Latest Max PC Tweets
- maximumpc: Original RickRoll Video Returns to YouTube After 24-Hour Copyright Hiatus http://t.co/MaH9Sxyy1 day 17 hours ago
- maximumpc: Corsair has decided not to go forward with its $78 million IPO, citing weak equity market conditions http://t.co/A5a4DAzj1 day 17 hours ago
- maximumpc: Want to work at Maximum PC? Our Online Managing Editor, Alex, is moving away, so we need a new one. http://t.co/9Z7JCh0E3 days 14 hours ago
