194 Environments of intelligence
machines – as, for example, reported by Pamela McCorduck (2004, Chapter 9)
with respect to Hubert Dreyfus’s critique of AI: if a machine can perform a certain
cognitive task, successful accomplishment of that task, whatever it is, by definition, cannot be thinking. Dreyfus himself states his critique as the problem of
the horizon of AI systems actually accomplishing the cognitive tasks they were
promised to accomplish to be “receding at an accelerating rate” because these
tasks turn out to be vastly more complex than AI researchers believed (1979, 5).
On either version of the argument, the boundary of what constitutes the accomplishment of a cognitive task will be continuously pushed upward with the advent
of more capable machines, and there is little we will have learned until we thereby
either ultimately carve out a domain for a Cartesian res extensa, or something to
the same effect, that would remain essentially inaccessible to any kind of machine
simulation – or stand to find out that nothing inaccessible is left.
What I have been suggesting in this book sought to turn this line of argument
on its head in order to put reasoning about human cognition on its feet: ever
more capable cognitive artefacts become part of human environments in ways
that highlight how we are coupled with our environments in perception, cognition
and action, and they do so by modifying the ways in which we are thus coupled.
These artefacts are not machines that think, nor are they supposed to be machines
that simulate human thinking. This may be the wrong accomplishment to seek
from them to begin with, particularly if the conception of the nature of cognition
that entered into that quest is at risk of being incorrect, inadequate or inaccurate.
Still, the abilities of cognitive artefacts move way beyond traditional conceptions
of what machines can do in a variety of practical ways that may be instructive to
inquiries into human cognition in different fashion. To end on an artefact-based
metaphor other than that of machines: they may serve as calibrating instruments,
in not simply measuring but helping to adjust our measurements of what sort of
environment-bound activity human thinking is.
machines – as, for example, reported by Pamela McCorduck (2004, Chapter 9)
with respect to Hubert Dreyfus’s critique of AI: if a machine can perform a certain
cognitive task, successful accomplishment of that task, whatever it is, by definition, cannot be thinking. Dreyfus himself states his critique as the problem of
the horizon of AI systems actually accomplishing the cognitive tasks they were
promised to accomplish to be “receding at an accelerating rate” because these
tasks turn out to be vastly more complex than AI researchers believed (1979, 5).
On either version of the argument, the boundary of what constitutes the accomplishment of a cognitive task will be continuously pushed upward with the advent
of more capable machines, and there is little we will have learned until we thereby
either ultimately carve out a domain for a Cartesian res extensa, or something to
the same effect, that would remain essentially inaccessible to any kind of machine
simulation – or stand to find out that nothing inaccessible is left.
What I have been suggesting in this book sought to turn this line of argument
on its head in order to put reasoning about human cognition on its feet: ever
more capable cognitive artefacts become part of human environments in ways
that highlight how we are coupled with our environments in perception, cognition
and action, and they do so by modifying the ways in which we are thus coupled.
These artefacts are not machines that think, nor are they supposed to be machines
that simulate human thinking. This may be the wrong accomplishment to seek
from them to begin with, particularly if the conception of the nature of cognition
that entered into that quest is at risk of being incorrect, inadequate or inaccurate.
Still, the abilities of cognitive artefacts move way beyond traditional conceptions
of what machines can do in a variety of practical ways that may be instructive to
inquiries into human cognition in different fashion. To end on an artefact-based
metaphor other than that of machines: they may serve as calibrating instruments,
in not simply measuring but helping to adjust our measurements of what sort of
environment-bound activity human thinking is.