(Report comment)

There is no qustion that America is less secure in the event of an actual war with either Russia or China than it was 50 years ago. Why did this happen? Does this history have any value today?

See my blog

http://warbyiq.com/ R. Peppe

(Report comment)

A computer is already self-aware (POST on boot for example) doing system checks and stuff…knowing where the computer is and how or what internet connection they can accept. All of that exists today. The only thing missing in the ‘terminator’ scenario is the ability to learn, sure we can use genetic algorithms to solve problems faster but fo a computer at this stage to take outside imput and to form their OWN IDEA…I am going to stree OWN IDEA… about the idea is what really we are talking about, the rest of the posturing is just that, computers are ’self aware’ and ’self correcting’ to a point now.

Get a computer to actually start to learn new skills without a developer progaming anyting and you got a nobel prize and probable SkyNet (joke).

(Report comment)

Sorry to say but all the your two statements above are already available, if you check out thehttp://en.wikipedia.org/wiki/Genetic_algorithm they actually discuss the topics you say can not happen in hardward and software. Genetic Algorithms have been able to approximate a number of things so I wouldn’t call it a joke if all your basing AI on is your couple of statements….AI is a lot more that what you mentioned.

(Report comment)

You wouldn`t need type face recognition unless you wanted them to scan stuff in…otherwise the ‘terminator bot’ can just grab everything lately off the internet.

(Report comment)

If, however, we begin to construct the basic hardware
that EMULATES our own organic neural net, then we’ll
get a machine that can rival and even exceed our own
capabilities.

Using an example, IBM fairly recently sumulated the
cortical region of a rat on a massively parallel
supercomputer called Blue Gene. By emulating the actual
molecular and chemical structure of that cortex in software,
the designers were able to run a 10 second simulation of
its operation and were thus able to perform ANY operation
that would normally be used in that brain structure
-albeit quite slowly-.

Applying such simulation of basic brain structure, we would
be able to DUPLICATE the general pattern matching capabilities
of humans within software, which then allows the simulation
to learn very much like we do. If we give the simulation
MULTIPLE hardware inputs such as many Internet connections,
multiple machine vision systems with pre-built font and
typeface recognition and audio inputs, the emulated human
brain could learn at a much more accelerated rate than humans.

By sending the emulated brain structure to “School” with
multiple teachers “teaching” it 24 hours a day, 365 days a
year in shifts, we could turn that un-organized neural net
from a baby into a graduate level student in probably less
than 4 years. AND it would aquire, from its internal pattern
matching abilities, a semblance of what we call a PERSONALITY.

The quality of that personality would be HIGHLY DEPENDENT
upon the initial conditions of it’s learning environment.
So the warning here is that a fully software-emulated
brain structure would need to be taught ALL the basics
of human ethics, morality, art, science, social etiquette
and politics by RESPONSIBLE, ETHICAL and DISCIPLINED people.

This would have an effect much like how we teach a child.
Children taught by ethical, moral and disciplined people
generally turn into ethical, moral and self-disciplined
adults. This final learning could then be copied to other
hardware to allows us a DEFENCE against the inevitable
ROGUE artificial intelligences that will be eventually
created by rogue nasty, unethical, immoral people.

The key here is who gets up and running a fully grown
AI first! — First one out of the gate usually wins!

The base hardware challenges is that some people believe
that a 100 Petaflop (i.e. 100 Quadrillion Floating Point
Operations Per Second) supercomputer with a 100 Exabyte
memory storage capacity (100 Quintillion bytes) is enough
hardware to simulate the entire human brain’s molecular
structure at faster-than-human speeds.

This type of hardware capacity (if using Moores Law)
is only about 20 years away since the fastest Cray
Roadrunner supercomputers already does 1.17 Petaflops.
Double that speed every 18 months and you get past that
magical threshold of 100 petaflops to equal human brainpower
in a little over 20 years which is right on schedule.

AND we could fit that sort of hardware in a
small building the size of a 4 bedroom house.

I should ask a pointed question, that since we know the
actual chemical biology of the brain pretty well,
who’s to know or say whether or not at some secret lab
say at the NSA (National Security Agency) that
a basic human brain isn’t already being emulated
on a CLUSTER of 1+ petaflop IBM Blue Gene/Q
supercomputers? At today’s prices, 100 such machines
would be about 10 BILLION dollars (i.e. $100 million a pop)
which is a drop in the bucket of the U.S.’s 2009
Department of Defence budget of $550 BILLION DOLLARS!

And who’s to say that emulated brain isn’t already working
on designing it’s future upgraded self to be fully installed
within mobile hardware that has arms, legs and sensory inputs?

Can we say TERMINATOR scenario!

(Report comment)

(Report comment)

(Report comment)

However, your point about impediments could be right on. A machine could possibly (although we can only speculate on how possible this is since we have no evidence to go on at this point) see us as an impediment to it’s survival and try to remove us because of it. Ironically though, this would be a very Darwinian action – a concept that we understand well and should be able to anticipate. So I would hope that we would be smart enough to put safeguards in place to ensure this could not happen. After all, history does show us that simply relying on gratitude hardly ensures you will receive it.

(Report comment)

(Report comment)