| |
Relative Advantages of Computer Programs, Minds-in-General, and the
Human Brain
© 2003 by Anand and Michael Anissimov
Knowledge of the comparative strengths and weaknesses
of computer programs, minds-in-general (including Artificial Intellects),
and the human brain is of paramount importance to rigorous analysis
of the benefits and risks of AI and human intelligence enhancement.
Since humans evolved for millions of years to model, outwit, and predict
the psychology of other humans, we lack the cognitive
support for accurate analysis of complex computer programs and future
AIs, entities we never encountered in our ancestral environment. Accordingly,
we have a tendency to ascribe characteristics to minds-in-general that
are actually unique to humans. This class of cognitive error is known
as "anthropomorphism". Anthropomorphism results in a widespread
tendency to pigeonhole future AIs as an awkward chimera between mechanical,
automatic computer programs such as Microsoft Word, and scheming, repressed
humans. Unlike computer programs or human beings, AIs will be an entirely
new class of mind; in no way obligated to behave in accordance with
our default intuitions, except insofar as they are programmed (or reprogram
themselves) that way. Considering that humans possess many design features
that derive directly from our adaptive history and biological constraints,
and that Artificial Intellects would not be created by the same underlying
process, nor would they be composed of the same materials, we have little
reason to anticipate that AIs in general would gravitate towards specific
humanlike characteristics.
For example, some thinkers have assumed that if human
beings "set a positive example" for AIs, then that will increase
the probability that such AIs will behave in psychologically balanced,
pro-social, or altruistic ways. But this is not necessarily the case.
The cognitive machinery underlying the human tendency to absorb norms
and morals from our peers and society is exquisitely complex and precisely
designed by evolution for specific methods of functionality. We have
no reason to believe such complex machinery would arise spontaneously
in AIs without being explicitly programmed, any more than we should
expect a tornado passing through a junkyard to construct a 747 airplane.
If we did choose to construct AIs with the tendency to absorb norms
from peers, we will be free to structure the absorption algorithm in
ways entirely different than it appears in Homo sapiens sapiens
- our particular method of moral absorption is one among many, and does
not represent any theoretical ideal of moral, psychological, or inferential
optimality. "Setting a good example" may be a good approach
to fostering morality in some AI designs, but in others it could
be totally useless.
Just as we should not expect AIs to have humanlike goals or morals by
default, we should not expect AIs to have humanlike levels of intelligence,
areas of specialty, or cognitive abilities. We must consider the probable
underlying hardware and inherent advantages or disadvantages AIs might
have relative to humans or computer programs. In an effort to encourage
more detailed analysis of these issues, following are three lists of
relative advantages between computer programs, minds-in-general, and
the human brain.
Advantages of Computer Programs over the Human Brain
The following advantages of computer programs over the
human brain do not necessarily apply to advantages of minds-in-general
over the human brain.
More design freedom, including ease of modification
and duplication; the capability to debug, re-boot, backup and attempt
numerous designs.
The ability to perform complex tasks without making
human-type mistakes, such as mistakes caused by lack of focus, energy,
attention or memory.
The ability to perform extended tasks at greater
serial speeds than conscious human thought or neurons, which perform
approx. 200 calculations per second. Computing chips (~2 GHz) presently
have a 10 million to one speed advantage over our neurons.
The in principle capacity to function 24 hours
a day, seven days a week, 365 days a year.
The human brain can not be duplicated or "re-booted,"
and has already achieved "optimization" through design by
evolution, making it difficult to further improve.
The human brain does not physically integrate
well, externally or internally, with contemporary hardware and software.
The non-existence of "boredom" when
performing repetitive tasks.
Advantages of the Human Brain
over Minds-in-General
Present AIs lack human general intelligence and
multiple years of real-world experience.
The computational capacity of the human brain is estimated
at 2 * 10^16, or 20 million billion calculations per second, which is
twenty times greater than the supercomputer Blue Gene's predicted achievement
of 10^15, or 1 million billion calculations per second, by 2005. However,
the human brain may not have a computational advantage over computers
for much longer. Ray Kurzweil,
for example, predicts that the computational capacity of the human brain
will be accomplished on supercomputers, or clustered systems, by 2010,
followed on personal computers by 2020.
The human brain has already achieved a high-level
of complexity and "optimization" through design by evolution,
and thus has proven functionality.
Advantages
of Minds-in-General over the Human Brain
The following are not advantages
of specific AI approaches, but rather advantages of minds-in-general
over the human brain.
An increased ability to acquire, retrieve, store and
use information on the Internet, which contains most human knowledge.
Lack of human failings that result from complex functional
adaptations, such as observer-biased beliefs or rationalization.
Lack of neurobiological features that limit human control
over functionality.
Lack of complexity that we have acquired from evolutionary
design, e.g., unnecessary autonomic processes and sexual reproduction.
The ability to advance on the design of evolution, which
is continually constrained by blindness, the requirement to maintain
preexisting design, and a weakness with simultaneous dependencies.
The ability to add more computational power to a particular
feature or problem. This may result in moderate or substantial improvements
to preexisting intelligence. (AI does not have an upper limit on computational
capacity; we do.) Note that the speed of computational power is predicted
to continually increase exponentially, and decrease exponentially in
cost, every 12-24 months, in accordance with Moore's Law.
The ability to analyze and modify every design level
and feature.
The ability to combine autonomic and deliberative processes.
The ability to communicate and share information (abilities,
concepts, memories, thoughts) at a greater rate and on a greater level
than us.
The ability to control what is and what is not learned
or remembered.
The ability to create new modalities that we lack, such
as a modality for code, which may improve the AI's programming ability-by
making the AI inherently native to programming - far beyond our own
(a modality for code may allow the AI to perceive its hardware machine
code, i.e. the language used to write the AI, and other abilities).
The ability to learn new information very rapidly.
The ability to consciously create, analyze, modify,
and improve abilities, concepts, or memories.
The ability to operate on computer hardware that has
powerful advantages over human neurons, such as the ability to perform
billions of sequential steps per second.
The capacity to self-observe and understand on a fine-grained
level that is impossible for us. AIs may have an improved capacity for
introspection and manipulation, such as the ability to introspect and
manipulate code, which would be the functional level comparable to human
neurons, which we can't think about or manipulate.
The most important and powerful capacity of minds-in-general
over the human brain is the ability to recursively self-encapsulate
and self-improve its intelligence. As a mind becomes smarter, the mind
can use its intelligence to improve its design, thereby improving its
intelligence, which may allow further improvements to its design, thus
allowing further improvements to its intelligence. It's unknown when
open-ended self-improvement may begin. A conservative assumption is
human-similar general intelligence; but it may begin before then, and
it is important to plan nonconservatively.
The advantages of minds-in-general, and self-improving AIs in particular,
is elaborated at greater length in third section of the paper "Levels
of Organization in General Intelligence", released by the Singularity
Institute for Artificial Intelligence in 2002.
General References
Kurzweil, Ray. 2001. The
Law of Accelerating Returns. KurzweilAI.net, March 7, 2001.
Singularity Institute. 2001. Creating
Friendly AI.
Singularity Institute. 2001. What
is Seed AI?
Voss, Peter. 2001. Why
Machines Will Become Hyperintelligent before Humans Do.
Yudkowsky, Eliezer. 2002. Levels
of Organization in General Intelligence. To appear in Advances
in Artificial General Intelligence, Goertzel and Pennachin, eds.
(Back to articles.)
|
|
|
|
