Consciousness

other words, Kismet has no Cartesian Theatre  – but then,
as we concluded in Chapter 4, we don’t either. Like Kismet,
we humans have a subsumption architecture. That is,
evolution kept whatever worked, dropped what did not,
and piled new routines on top of old ones in haphazard
interacting layers without an overall plan. This is often how
robots improve as a team of scientists works on them.

All these robots are much simpler than us, but let’s
imagine some fanciful future descendants of today’s
social robots who are even more skilful. Imagine CREEPI
(Conscious Robot with Evolved Emotional and Phenom-
enal Intelligence), who is still just a mass of metal limbs,
motors, and chips, but has soft human-like skin that can
convey subtle facial expressions, and eyes that cry wet
tears accompanied by convincing sobs, activated by systems that respond to
the person who is in front of it. Imagine that CREEPI can respond to emotions
displayed by a human: laughing when the human laughs, or sympathising and
comforting someone who appears upset. Imagine that CREEPI is even more sen-
sitive to other people’s emotions than most humans. What would you say now?
Would you still be sure that CREEPI is just a pile of bits, or would you think that
maybe it was conscious?

An obvious response is this. We know that simple systems can mislead us into
thinking they have plans, goals, and beliefs when they haven’t, and that more
complex ones can mislead us even more. So we should not be fooled. We should
conclude only that CREEPI acts as if it is conscious, when it is not really conscious.

An alternative response is this. There is no dividing line between as-if and real
consciousness. Being able to sympathise with others and respond to their emo-
tions is one part of what we mean by consciousness. Today’s social robots have
a little bit of it and CREEPI has a lot. CREEPI is not conscious in the way that we
are because it is a social machine without other abilities, but within its limited
domain its consciousness is as real as any. Maybe, indeed, our own complexity is
what misleads us into believing that consciousness can increase only with grow-
ing complexity.

Which is right? And how can we find out?

uncanny va{ lley

healthy

person

humanoid robot

industrial

robot

bunraku puppet

stuffed

animal

corpse

zombie

prosthetic

hand

human likeness 50% 100%

familiarity

+

moving

still

FIGURE 12.19 • The ‘uncanny valley’ is the

name given to the dip in a

hypothetical graph plotting

emotional reactions against

the similarity of an artefact to

a human. The most negative

response occurs to robots or

toys that are ‘almost human’.

Movement increases the effect.

Chalmers, D. (1996). Strong artificial intelligence.

In D. Chalmers, The conscious mind: In search of a

fundamental theory (pp. 313–322). New York: Oxford

University Press.

Defends strong AI against objections, including the

Chinese Room.

READING