bit-tech.net

The missing component for true AI

Posted on 24th Aug 2009 at 10:17 by Mark Mackay with 18 comments

Mark Mackay
For years now the debate as to whether it’s possible to create an artificial intelligence by mimicking the human brain has raged. Last month, New Scientist published an awesome feature titled Memristor Minds, which has some interesting points to make about the AI debate.

The article tells the story of a chap named Leon Chua who, back in the early seventies, was an electronics engineer at the University of California, Berkley. He found it fascinating that no one had sat down and written a mathematical formula for electronics. Armed with his technical background and enthusiasm for mathematics, he was a fine candidate to have a stab at it. And that’s precisely what he did.

I’m not an electronics engineer so I’m going to quote New Scientist directly for this bit to make sure that I get it right.

" Back in 1971, Chua was examining the four basic quantities that define an electronic circuit. First, there is electric charge. Then there is the change in that charge over time, better known as current. Currents create magnetic fields, leading to a third variable, magnetic flux, which characterises the field's strength. Finally, magnetic flux varies with time, leading to the quantity we call voltage.

"Four interconnected things, mathematics says, can be related in six ways. Charge and current, and magnetic flux and voltage, are connected through their definitions. That's two. Three more associations correspond to the three traditional circuit elements. A resistor is any device that, when you pass current through it, creates a voltage. For a given voltage a capacitor will store a certain amount of charge. Pass a current through an inductor, and you create a magnetic flux. That makes five. Something missing?"


The missing component for true AI *The Missing Component For True AI?
This image needs no caption

Now, in mathematics there’s a rule which says that four interconnected variables can relate to one another in six different ways. When Chua started running the numbers, he found that there was something missing. Only five relations could be made using what we knew about electronics at the time. In a thought process the likes of which someone with my humble understanding of electronics could only muster after a good deal of study, Chua deduced that the sixth relationship would be a type of resistor that could some how remember the voltage that had previously passed through it. The Memristor was born.

Despite theory pointing towards the existence of the Memristor, at that point, no-one had actually made one. In the early 2000s, Leon Chua had the chance to pair up with Stan Williams, a researcher for Hewlett Packard, things started to get moving. The duo found a way to pass current through a semi-conductor called titanium dioxide in such a way that its resistance was altered. Crucially, when the current was switched off, the state of resistance at the time stayed the same and so the metal remembered and ‘woke up’ in the same resistance state.

While writing up his first paper on memristance, the ever-inquisitive Chua became fascinated with synapses in the human brain. While studying them, the electronics engineer realised synapses were beginning to appear startlingly familiar. He realised then that synapses were in fact, 'biological memristors’.

Research has already been begun by organisations such as DARPA to put these findings to work in the AI field. It’s starting to look as though the reason we’ve been so far from recreating an artificial ‘brain’ is because we may have been missing an artifial alternative to one of the most crucial components. We may have just taken a massive step towards making Jonny Five come alive.

18 Comments

Discuss in the forums Reply
alpaca 24th August 2009, 10:41 Quote
after the descrete graphics card, the discrete soundcard, now comes the discrete AI card. a bunch of those memristors on a pcb and you can play agianst a real challenge! no need for other people anymore!
Combinho 24th August 2009, 10:43 Quote
A very interesting and informative blog.
Tyrmot 24th August 2009, 11:25 Quote
I read that NS article, it was very good. However I think there's still a lot more to come in developing true AI than a missing part - everyone always seems to say it's 'just around the corner' and then some new level of complexity is discovered which pushes it back again. Rather than designing AI in a sort of 'reverse-engineersing the human brain' kind of way, perhaps the way to do it is simply to attempt to arrange matters so that it can arise from the underlying architecture? Is that what is happening in this field? Wish I knew more about it really... definitely one of the most interesting areas of IT today, and receives less attention than it deserves
LAGMonkey 24th August 2009, 11:42 Quote
skynet?!
Paradigm Shifter 24th August 2009, 11:53 Quote
If anyone wants to read a fascinating paper on memristors, here's the reference:

Strukov, D.B., Snider, G.S., Stewart, D.R., Williams, R.S. (2008) The Missing Memristor Found. Nature. 453(1): 80-3.

Shows a lot of promise for super-high-density low-volatility memory.
Cutter McJ1b 24th August 2009, 11:54 Quote
Quote:
Originally Posted by Tyrmot
I read that NS article, it was very good. However I think there's still a lot more to come in developing true AI than a missing part - everyone always seems to say it's 'just around the corner' and then some new level of complexity is discovered which pushes it back again. Rather than designing AI in a sort of 'reverse-engineersing the human brain' kind of way, perhaps the way to do it is simply to attempt to arrange matters so that it can arise from the underlying architecture? Is that what is happening in this field? Wish I knew more about it really... definitely one of the most interesting areas of IT today, and receives less attention than it deserves

From what I gathered researching for the Artificial Intelligence feature that we wrote, it seemed to me more like we had hit something of a deadend in our searh for creating a true AI rather than it being just round the corner. There's more to conciousness than just a hardware and software combination and that's all we can do. The memristor may change that provide us a way out and set us on a path to creating a real Jonny Five. However there also may be more pieces of the puzzle that are yet to be discovered.
Hugo 24th August 2009, 14:07 Quote
Okay, so basically Leon Chua and Stan Williams are the guys who we need to go back in time and kill to stop Skynet being created?
Cutter McJ1b 24th August 2009, 14:11 Quote
Quote:
Originally Posted by HugoB
Okay, so basically Leon Chua and Stan Williams are the guys who we need to go back in time and kill to stop Skynet being created?

You win the reading-between-the-lines-of-the-blog competition. That is precisely the case. Work is currently under way to finish the time machine, though attempts are in jeopardy of being foiled by evil robots that have been sent back in time by Skynet. Watch this space.
thehippoz 24th August 2009, 15:23 Quote
Quote:
Originally Posted by Cutter McJ1b
From what I gathered researching for the Artificial Intelligence feature that we wrote, it seemed to me more like we had hit something of a deadend in our searh for creating a true AI rather than it being just round the corner. There's more to conciousness than just a hardware and software combination and that's all we can do. The memristor may change that provide us a way out and set us on a path to creating a real Jonny Five. However there also may be more pieces of the puzzle that are yet to be discovered.

hehe yeah it's crap to mystify the masses.. so what, it holds it's 'memory' or resistance in a powered off state- it's old tech too.. how is that any better than ram when it comes to the task of creating true ai- imo the more they go down this route, the more fail your gonna have

computers just do exactly what they are programmed to do.. nothing more, nothing less.. until they can somehow master biology- that's a long way off imo.. we won't see a true ai, in reality all these guys trying to create a working ai with hardware, if they step back.. on/off that's all it is, it maybe complex but nothing intelligent

we may never figure it out :D
Cutter McJ1b 24th August 2009, 15:59 Quote
Quote:
Originally Posted by thehippoz
we may never figure it out :D

Yup, it's true, we may never figure it out. But. If we don't try then we will definitely never figure it out. If you want a comprehensive answer as to why a memristor is different to RAM in the task of creating an AI, let's revert back to New Scientist's feature.

'In true memristive fashion, Chua had anticipated the idea that memristors might have something to say about how biological organisms learn. While completing his first paper on memristors, he became fascinated by synapses - the gaps between nerve cells in higher organisms across which nerve impulses must pass. In particular, he noticed their complex electrical response to the ebb and flow of potassium and sodium ions across the membranes of each cell, which allow the synapses to alter their response according to the frequency and strength of signals. It looked maddeningly similar to the response a memristor would produce. "I realised then that synapses were memristors," he says. "The ion channel was the missing circuit element I was looking for, and it already existed in nature."
The behaviour of synapses looked maddeningly similar to a memristor's response

To Chua, this all points to a home truth. Despite years of effort, attempts to build an electronic intelligence that can mimic the awesome power of a brain have seen little success. And that might be simply because we were lacking the crucial electronic components - memristors.'
Xtrafresh 25th August 2009, 07:26 Quote
Quote:
Originally Posted by thehippoz
computers just do exactly what they are programmed to do.. nothing more, nothing less.. until they can somehow master biology- that's a long way off imo..
What is this "biology" you speak of? Do you mean a bag of complicatedly managed chemicals? Computers can do that. Do you mean evolution? Computers can do that too, i submit to you that the current generations of computers would have been impossible to build without the last.
RotoSequence 25th August 2009, 09:43 Quote
If the patterns of thought can be mimicked with hardware and software, AI is possible. Since we've been simulating small parts of the brain with brute force software techniques and getting similar results as biological brains, it's probably just a matter of time.
Cutter McJ1b 25th August 2009, 10:07 Quote
Quote:
Originally Posted by RotoSequence
Since we've been simulating small parts of the brain with brute force software techniques.....

Do you have a link to any good examples of this?
TSR2 25th August 2009, 11:49 Quote
I can see this being pretty important, isn't the brain different from a machine in that it can think in shades of grey, not just 0's and 1's. Though I can't see myself buying a robotic, artificially intelligent, servant soon
Star*Dagger 25th August 2009, 19:35 Quote
2 comments

1) the brain is holographic, you do it all or you do it wrong.

2) AI will not be in computer games until there is no more Ministry or Department of Defense. The US military already has cloaking, you do not see that...

Yours in harsh but True Plasma,
Star*Dagger
OtakuHawk 25th August 2009, 19:49 Quote
Quote:
Originally Posted by alpaca
no need for other people anymore!

that is kind of what we're worried about.
RotoSequence 26th August 2009, 21:33 Quote
Quote:
Originally Posted by Cutter McJ1b
Do you have a link to any good examples of this?

http://bluebrain.epfl.ch/

The BBC covered the story as well;

http://news.bbc.co.uk/2/hi/science/nature/8012496.stm
BLC 8th September 2009, 13:27 Quote
Extremely interesting read. It's going to take years of research to even try to see what practical implications any of this will have, but isn't that what science is all about?
Log in

You are not logged in, please login with your forum account below. If you don't already have an account please register to start contributing.



Discuss in the forums
Fractal Design Node 804 Review

Fractal Design Node 804 Review

17th September 2014

SK Hynix SH910A SSD 256GB Review

SK Hynix SH910A SSD 256GB Review

16th September 2014

1200W PSU Roundup 2014

1200W PSU Roundup 2014

15th September 2014