Form vs function
8 views
Skip to first unread message
* Glenn Becker
Dec 29, 2010, 1:54:28 PM12/29/10
to The ALISA Community
The Dec 2010 issue of IEEE Spectrum includes an article "MoNETA: A
Mind Made from Memristors" by M Versace and B Chandler
The article can be seen here:
http://spectrum.ieee.org/robotics/artificial-intelligence/moneta-a-mind-made-from-memristors/?utm_source=techalert&utm_medium=email&utm_campaign=101202
The authors argue the following:
"Researchers have suspected for decades that real artificial
intelligence can't be done on traditional hardware, with its rigid
adherence to Boolean logic and vast separation between memory and
processing."
They argue, it seems, on behalf of DARPA's Grand Challenge and HP the
memristor manufacturer that the form is critical to the function.
I disagree. First of all, I don't believe that AI must necessarily
mimic biological intelligence. Nature had amino acids and Na ions to
work with and arrived at one solution. I don't believe a Si-based
machine needs exactly the same form.
Sure we talk a lot about biological precedent, but at what level(s)
must form match?
I believe that the drives (or motivations) we bestow upon AI entities
will have a much greater effect on the resulting intelligence than the
type of hardware. Sure, higher density and lower power is better and
to the extent that memristors can provide them, Great. But I think
there is a lot of potential in decentralized network-based
intelligence as well, that will certainly be very different from
biological intelligence at the neuron level but more similar at the
brain level.
Thoughts?
Mind Made from Memristors" by M Versace and B Chandler
The article can be seen here:
http://spectrum.ieee.org/robotics/artificial-intelligence/moneta-a-mind-made-from-memristors/?utm_source=techalert&utm_medium=email&utm_campaign=101202
The authors argue the following:
"Researchers have suspected for decades that real artificial
intelligence can't be done on traditional hardware, with its rigid
adherence to Boolean logic and vast separation between memory and
processing."
They argue, it seems, on behalf of DARPA's Grand Challenge and HP the
memristor manufacturer that the form is critical to the function.
I disagree. First of all, I don't believe that AI must necessarily
mimic biological intelligence. Nature had amino acids and Na ions to
work with and arrived at one solution. I don't believe a Si-based
machine needs exactly the same form.
Sure we talk a lot about biological precedent, but at what level(s)
must form match?
I believe that the drives (or motivations) we bestow upon AI entities
will have a much greater effect on the resulting intelligence than the
type of hardware. Sure, higher density and lower power is better and
to the extent that memristors can provide them, Great. But I think
there is a lot of potential in decentralized network-based
intelligence as well, that will certainly be very different from
biological intelligence at the neuron level but more similar at the
brain level.
Thoughts?
* Glenn Becker
Dec 30, 2010, 10:39:10 PM12/30/10
to alisa-c...@googlegroups.com
--- emailed by Michael LeChat 12/29/2010
"mimicing biological intelligence"--you are right here. As long as we specify what kind of substate we are dealing with. The structural isomorphism of biological entities, and their consequent capacities, is part of evolutionary theory, i.e. homologs must be in direct line of descent and therefore (roughly) morphologically similar. Not necessarily so with AI, save possibly with regard to emotion/pain issues. Nice debating point, Glenn!
John
Jan 14, 2011, 1:01:13 PM1/14/11
to * Glenn Becker, The ALISA Community
Hi Glenn,
I think I agree with just about everything you're saying here.
You're point re AI not following bio intelligence couldn't be much more true. We use semiconductors and integrated circuitry to increase the speed and number of gates. A CPU is an analog to a brain of sorts but they have vastly different forms. Even though the CPU form is not natural, its function supernatural in ways. We've demonstrated some amazing use of lot of fast gates when you consider Deep Blue and now Watson - Watson now poised to compete against Jeopardy champions.
I further agree regarding the drives we will bestow upon such entities. Our direct instructions will be independent of the minor variations among machines they can be carried out on.
I'd be keen to hear other thoughts on this as well!
John Schuster
I think I agree with just about everything you're saying here.
You're point re AI not following bio intelligence couldn't be much more true. We use semiconductors and integrated circuitry to increase the speed and number of gates. A CPU is an analog to a brain of sorts but they have vastly different forms. Even though the CPU form is not natural, its function supernatural in ways. We've demonstrated some amazing use of lot of fast gates when you consider Deep Blue and now Watson - Watson now poised to compete against Jeopardy champions.
I further agree regarding the drives we will bestow upon such entities. Our direct instructions will be independent of the minor variations among machines they can be carried out on.
I'd be keen to hear other thoughts on this as well!
John Schuster
Reply all
Reply to author
Forward
0 new messages