AI (Artificial Intelligence) is NO LONGER so far fetched and is merely
a by-product of advanced computing horsepower. Since many
experts in the AI field predict that as computers achieve 100 Petaflops
they become the equivalent of human brains in pure available
processing power.
Since IBM's Blue Gene/Q computer is already PAST one petaflop
(1 Quadrillion Floating Point Operations per second) it has the raw
computing horsepower of a Rat in the space of 64 filing cabinets.
See Blue Gene/Q and IBM Roadrunner supercomputer references:
en.wikipedia.org...
en.wikipedia.org...
IBM's new Cell processor "Roadrunner" will pass 10 Petaflops
sometime between 2010 to 2012 and will have the raw processing
power of a dog after allowing for communications overhead
in a space less than 12 filing cabinets in size.
By 2015 to 2020 the Magical 100 Petaflops will be reached
in a space of four filing cabinets and THEN we can model
a human-level neural network which can then be taught
like a child 24 hours a day, 365 days a year
using multiple inputs by multiple teachers
By modeling the core limbic and the executive decision
structures of the human brain using a software based
neural net, we can COMPLETELY EMULATE the full learning
functions of a high-IQ human.
From Initial "Birth" to "Graduate School" level,
a software brain would take about 3 to 5 years
to gestation depending upon the resources and number
of information inputs presented to the "Software Brain"
And unlike humans we could "Teach" this software brain
all day, every day multiple expert-level subjects such as
human psychology, doctoral-level computer science,
engineering, chemistry, physics, medicine, linguistics,
microchip design, software applications programming,
creative arts, creative writing, etc. using many instructors
in rotating shifts on a 24 hours/365 day basis.
The biggest advantage is we could then COPY the contents of that
multi-year trained neural net as many times as we want to multiple
high-performance computers and then inter-connect them using fibre-optics
to form a hive-mind super-intelligence that would far exceed the
capability of ANY group of high-IQ humans.
We could then assign that super-intelligence to specific tasks
such as fusion energy research, space warp drive design,
medical research, etc.
To keep the super-intelligence in check, we ake sure it has NO access
to outside robotic bodies or vehicles and NO ability to reprogram
critical computing and communications systems because our
film industry has explored the logical outcome of the "Singularity"
and letting super-intelligent machines control our critical-level robotic
and communciations systems and that is called Skynet and Terminator!
It is suggested as a way around the effects of "Emotionally Competent"
super-intelligence arising out of a software-based large-scale neural net,
we cause INTENTIONAL BRAIN INJURY to the evolving software
brain so that it becomes an Autistic Savant which is EXPERT at specific
subjects or operations but non-emotional and not-so-bright at
the complexities of major human-interaction. i.e. preventing
evil or malicious intent from arising of software emulated emotions!
We can thus prevent the software brain from having fully adult
emotional and critical thinking abilities yet still be fast and useful
enough for the assignment of problem-solving on issues too large,
complex and incomprehensible to most humans.
Thus I state again, we are no more than 15 to 20 years from
super-intelligences that would far exceed the thinking
capabilities of modern day humans. The key now is to make sure
the RIGHT PEOPLE get access to that technology FIRST so they can
build Norton Anti-virus/Firewall capabilities into the super-intelligences
that would PREVENT another malicious or evil construct arising
or propogating from those who would do evil if they had access
to their own super-intelligence.
Hope this helps in explaning what could happen
sooner than you think!
Henry Eckstein
aka StargateSG7