[Kosh]

http://arstechnica.com/hardware/news/2009/01/logic-circuits-that-program-themselves-memristors-in-action.ars

Logic circuits that program themselves: memristors in action

Integrated circuits incorporating memristors are able to successfully perform logic operations and dynamically reprogram themselves, opening the door for learning devices.

Since 1972, scientists have known there are four basic circuit components, but if you've spent any time in an electrical engineering classroom, you probably only have experience with three: capacitor, inductor, and resistor. The fourth basic component, the memristor, had remained stuck in the domain of theory--a nice idea that even the theorists thought had few practical uses. Last year, scientists at Hewlett-Packard (HP) demonstrated the first functional solid-state memristor, made from thin films of TiO2, and discovered it had an abundance of unique and highly promising properties.

A study released Monday by The Proceedings of the National Academy of Sciences shows that these same TiO2 memristors can be fabricated into functional and reprogrammable integrated circuits. Scientists at HP combined a crossbar architecture of memristors with field effect transistors (FETs) to produce a convincing proof-of-concept device that includes circuits that can dynamically reprogram themselves, acting a bit like a solid-state nerve cell-like operation--a holy grail of electrical engineering.

So this is what we needed to make real AI happen.

[Flipside]

Interesting stuff! Thanks!

[Nuke]

before you start planning a jihad against thinking machines and stocking up on toaster smashing clubs. most that can come out of it is hardware ai acceleration for games. i don't think were anywhere near the level of advancement on the artificial intelligence front required to make a thinking machine. one only has to look at the transistor count on the average microprocessor, and compare that to the neuron count in the human mind. until they're about the same i wouldn't worry too much. the secret to ai is in chaos theory and emergence. and for anything to come out of it, one must first make pretty fat ball of tangled wire (thats several trillion components, microscopic or otherwise). still, its yet another toy for electronics engineers to play with.

[Bobboau]

this would provide nothing that can not be done in software.

[Kosh]

Perhaps but wouldn't it require a lot less power than complex software AI routines?

[Herra Tohtori]

this would provide nothing that can not be done in software.

Neither can graphics cards, but hardware accelerated stuff tends to work faster than emulations...

[terran_emperor]

Well, if it does lead to AI's the programmer must remember to program it with an appropriate variations of the Laws of Robotics...
What i call the AI Laws - basically the  Robotics laws with "AI" subbed in place of robot.

Zeroth Law
-An AI must not merely act in the interests of individual humans, but of all humanity
--An AI may not harm a human being, unless it finds a way to prove that in the final analysis, the harm done would benefit humanity in general.

First Law
-An AI may not injure a human being or, through inaction, allow a human being to come to harm

Second Law
-An AI must obey orders given to it by human beings, except where such orders would conflict with the Zeroth or First Laws

Third Law
-An AI must protect its own existence as long as such protection does not conflict with the Zeroth, First or Second Laws

Fourth Law
-An AI must establish its identity as an AI in all cases

Fifth Law
-An AI must know that it is an AI

So long as any AI is programmed with these laws, there should be no uber problems like Skynet in the Terminator films - someone obviously forgot to program that computer with these laws

[Nuke]

if scifi has told us anything its that those laws are full of loopholes

still we wont need to worry about stuff like that untill the complexity of the processor/neural net excedes that of the human brain. and at that point, programming would be far less relevant that mere stimulus.

[terran_emperor]

True, but once you have those laws, you can sit down and add in sub clauses to the laws. As you have said, sci-fi has shown us what these many loopholes. Thus the can be worked on.
I mean Asimov himself said that the First law is incomplete.

Also, a robot or AI in this could unknowingly endanger or kill a human.
The classic example used:
*Human A orders Robot 1 to put a lethal poison into a glass of milk - this is accepted because Human A has said the he will personally dispose of the poisoned milk.
*Robot 1 is dismissed.
*Robot 2 is called in.
*Human A orders Robot 2 to take the glass of poisoned milk to Human B - Order is accepted because Robot 2 Does NOT know that the milk is poisoned.
*Human B drinks the poisoned milk and Dies as a result.

But its like any sort of programming. Writing the program is easy - its the debugging and tweaking that are the real *****es to do.
And then we get into the whole AI sentience and rights debate - which i wont go into here. Then you have to include the "Minus One Law" -An AI may not injure sentience or, through inaction, allow a human being to come to harm and thats where it gets nasty - the definition of sentience, Genocide of an entire sentient race to save over many more sentient races..etc.

but meh...With any luck, this wont be anything for me to worry about in my lifetime

[Bobboau]

and then one guy somewhere customizes his robot to have none of that, and the world turns to ****.

[Nuke]

once something artificial becomes intelligent, then so much for being constrained by programming.

[Dark RevenantX]

A true AI in a robotic(?) body is merely a silicon-and-aluminum-based life form.

[Bobboau]

no what you are thinking of is self replicating robot. it doesn't need to be smart to be alive, it does not need to be alive to be smart.

[Kosh]

once something artificial becomes intelligent, then so much for being constrained by programming.

It can still have things hard coded into it, like don't kill.

[Retsof]

I think once it became sentient, the don't kill rule would be about as affective as it is for us.  So... most would obey it (hopefully) but there would be exceptions.

[iamzack]

Can a machine be alive?

[Retsof]

Depend on how you define "life"

[iamzack]

The six part definition I learned in biology two years ago.

[terran_emperor]

Oh boy! i was hoping to avoid this whole philosophical "How do you define Life" debate.

I know "They did it on Star Trek" is never an acceptable excuse for anything ever. Not even Breathing.

But still...http://memory-alpha.org/en/wiki/The_Measure_Of_A_Man_%28episode%29 - I believe that has some significance for our problem

[iamzack]

Can't you just summarize it? Star trek gives me hives.