Moravecs paradox

Moravecs paradox is the discovery by articial intelintrinsically dicult; it just seems so when we
ligence and robotics researchers that, contrary to tradido it.[3]
tional assumptions, high-level reasoning requires very little computation, but low-level sensorimotor skills require A compact way to express this argument would be:
enormous computational resources. The principle was
articulated by Hans Moravec, Rodney Brooks, Marvin
We should expect the diculty of reverseMinsky and others in the 1980s. As Moravec writes, it is
engineering any human skill to be roughly
comparatively easy to make computers exhibit adult level
proportional to the amount of time that skill has
performance on intelligence tests or playing checkers, and
been evolving in animals.
dicult or impossible to give them the skills of a oneyear-old when it comes to perception and mobility.[1]
The oldest human skills are largely unconscious and
so appear to us to be eortless.
Marvin Minsky emphasizes that the most dicult human skills to reverse engineer are those that are uncon Therefore, we should expect skills that appear efscious. In general, we're least aware of what our minds
fortless to be dicult to reverse-engineer, but skills
do best, he writes, and adds we're more aware of simple
that require eort may not necessarily be dicult to
processes that don't work well than of complex ones that
engineer at all.
work awlessly.[2]
Some examples of skills that have been evolving for millions of years: recognizing a face, moving around in
1 The biological basis of human space, judging peoples motivations, catching a ball, recognizing a voice, setting appropriate goals, paying attenskills
tion to things that are interesting; anything to do with perception, attention, visualization, motor skills, social skills
One possible explanation of the paradox, oered by
and so on.
Moravec, is based on evolution. All human skills are implemented biologically, using machinery designed by the Some examples of skills that have appeared more reprocess of natural selection. In the course of their evolu- cently: mathematics, engineering, human games, logic
tion, natural selection has tended to preserve design im- and much of what we call science. These are hard for
provements and optimizations. The older a skill is, the us because they are not what our bodies and brains were
more time natural selection has had to improve the de- primarily evolved to do. These are skills and techniques
sign. Abstract thought developed only very recently, and that were acquired recently, in historical time, and have
consequently, we should not expect its implementation to had at most a few thousand years to be rened, mostly by
cultural evolution.[lower-alpha 1]
be particularly ecient.
As Moravec writes:

2 Historical inuence on articial

intelligence

Encoded in the large, highly evolved sensory and motor portions of the human brain is
a billion years of experience about the nature
of the world and how to survive in it. The deliberate process we call reasoning is, I believe,
the thinnest veneer of human thought, eective only because it is supported by this much
older and much more powerful, though usually
unconscious, sensorimotor knowledge. We are
all prodigious olympians in perceptual and motor areas, so good that we make the dicult
look easy. Abstract thought, though, is a new
trick, perhaps less than 100 thousand years old.
We have not yet mastered it. It is not all that

In the early days of articial intelligence research, leading

researchers often predicted that they would be able to create thinking machines in just a few decades (see history of
articial intelligence). Their optimism stemmed in part
from the fact that they had been successful at writing programs that used logic, solved algebra and geometry problems and played games like checkers and chess. Logic
and algebra are dicult for people and are considered a
sign of intelligence. They assumed that, having (almost)
solved the hard problems, the easy problems of vision
and commonsense reasoning would soon fall into place.
1

6 BIBLIOGRAPHY

They were wrong, and one reason is that these problems

are not easy at all, but incredibly dicult. The fact that
they had solved problems like logic and algebra was irrelevant, because these problems are extremely easy for
machines to solve.[lower-alpha 2]
Rodney Brooks explains that, according to early AI research, intelligence was best characterized as the things
that highly educated male scientists found challenging,
such as chess, symbolic integration, proving mathematical theorems and solving complicated word algebra problems. The things that children of four or ve years could
do eortlessly, such as visually distinguishing between a
coee cup and a chair, or walking around on two legs, or
nding their way from their bedroom to the living room
were not thought of as activities requiring intelligence.[4]
This would lead Brooks to pursue a new direction in
articial intelligence and robotics research. He decided
to build intelligent machines that had No cognition. Just
sensing and action. That is all I would build and completely leave out what traditionally was thought of as the
intelligence of articial intelligence.[4] This new direction, which he called "Nouvelle AI" was highly inuential
on robotics research and AI.[5]

See also
Embodied philosophy
Embodied cognition
Nouvelle AI
Subsumption architecture
Hans Moravec
History of articial intelligence

Notes

[1] Even given that cultural evolution is faster than genetic

evolution, the dierence in development time between
these two kinds of skills is ve or six orders of magnitude, and (Moravec would argue) there hasn't been nearly
enough time for us to have mastered the new skills.
[2] These are not the only reasons that their predictions did
not come true: see the problems.

References

[1] Moravec 1988, p. 15.

[2] Minsky 1988, p. 29.
[3] Moravec 1988, pp. 1516.

[4] Brooks (2002), quoted in McCorduck (2004, p. 456)

[5] McCorduck 2004, p. 456.

6 Bibliography
Brooks, Rodney (1986), Intelligence Without Representation, MIT Articial Intelligence Laboratory
(2002), Flesh and Machines, Pantheon
Books
Campbell, Jeremy (1989), The Improbable Machine, Simon and Schuster, pp. 3031
Minsky, Marvin (1986), The Society of Mind, Simon
and Schuster, p. 29
Moravec, Hans (1988), Mind Children, Harvard
University Press
McCorduck, Pamela (2004), Machines Who Think
(2nd ed.), Natick, MA: A. K. Peters, Ltd., ISBN 156881-205-1, p. 456.
Nilsson, Nils (1998), Articial Intelligence: A New
Synthesis, Morgan Kaufmann, p. 7, ISBN 978-155860-467-4

Text and image sources, contributors, and licenses

7.1

Text

Moravecs paradox Source: https://en.wikipedia.org/wiki/Moravec{}s_paradox?oldid=659139375 Contributors: Vroman, AnonMoos,

Neilc, Mindmatrix, Wbeek, Vegaswikian, SmackBot, Simeon, Gregbard, Was a bee, Jamelan, CharlesGillingham, Retama, SchreiberBike,
Addbot, Tassedethe, Luckas-bot, Citation bot, Citation bot 1, EleferenBot, ZroBot, ClueBot NG, Helpful Pixie Bot, Jochen Burghardt,
Lgfcd, Omeglegirlslivecha and Anonymous: 21

7.2

Images

7.3

Content license

Creative Commons Attribution-Share Alike 3.0