Ants and robots, parlour games and steam drills 3
Or is it an advantage? Both AI and its philosophical critiques to a large extent
have been and remain living under the spell of the notion of the Turing Test, which
essentially depends on drawing that “fairly sharp line”. Taking this line to be a
firm boundary rather than a heuristic, that test evolved into something quite far
removed from Alan Turing’s classical “imitation game”, as described in the previous quote. It is the notion of such a firm boundary that has attracted criticism from
the camp of 4E cognition with some justification and in some detail.
However, as has been argued with some justification, too (most prominently
by Copeland 2000; Moor 1976; Whitby 1996), the aim of Turing’s thoughtexperimental game of machines imitating human conversational behaviour was
not to prove that machines could think, or to provide a test for whether they can
think. After all, the “imitation game” was playfully fashioned after an eponymous
popular Victorian parlour game.
2
Instead, that game was part of Turing’s inquiry
into the possible scope and depth of the tasks he designed for a certain class of
theoretical machines when devising his theory of computability in Turing (1936).
He considered digital computers one possible material incarnation of those theoretical machines – and human computers another, which served as the blueprint
for the former.
3
Turing’s theory of computability was driven by a genuinely meta-mathematical
interest: if, within the confines of a logical calculus, there is an unequivocal, welldefined and finite, hence at least in principle executable procedure for deciding on
the provability of a proposition that has been stated in that calculus, this procedure
should be translatable into arithmetical forms. These, in turn, could be broken
down into a set of simple mathematical routines that would be executable for a
human ‘computer’, that is a person with basic mathematical skills who is provided
with a set of input numbers and a set of instructions and then ordered to calculate
the result (and pass on that result as input for further computation). These inputs
and instructions, Turing’s argument continues, would be simple and unequivocal
enough to be handled by a suitably designed machine, too.
Although representing only an expressly restricted subset of human intellectual abilities, these tasks were well beyond the scope of traditional, cog-beltand-pulley machines. Margaret Boden (2006, 168f ) notices that the thought that
machines could possibly think was not even a ‘heresy’ up to the early twentieth
century, as that claim would have been all but incomprehensible. The reason is
that intellectual capabilities were thought of as essentially tied to the (most likely)
human organism whereas the capabilities of machines were mostly thought of
as being restricted to the exertion and harnessing of physical forces. More precisely, according to machine-age definitions of machines, “the primary function
of a mechanical device can be either the modification of motion (direction) or the
modification of motion and force (amplification and reduction)”, where the former
would be a mechanism and the latter a machine proper (Mitcham 1994, 170, in his
reconstruction of definitions of machines that go back to the mid-nineteenth century). A more reductive definition confines the activities of machines to “changing
the direction of motion” of matter, because “moving matter is all the force with
which machines deal” (“What is a Machine?” 1872, 39). A more differentiated
Or is it an advantage? Both AI and its philosophical critiques to a large extent
have been and remain living under the spell of the notion of the Turing Test, which
essentially depends on drawing that “fairly sharp line”. Taking this line to be a
firm boundary rather than a heuristic, that test evolved into something quite far
removed from Alan Turing’s classical “imitation game”, as described in the previous quote. It is the notion of such a firm boundary that has attracted criticism from
the camp of 4E cognition with some justification and in some detail.
However, as has been argued with some justification, too (most prominently
by Copeland 2000; Moor 1976; Whitby 1996), the aim of Turing’s thoughtexperimental game of machines imitating human conversational behaviour was
not to prove that machines could think, or to provide a test for whether they can
think. After all, the “imitation game” was playfully fashioned after an eponymous
popular Victorian parlour game.
2
Instead, that game was part of Turing’s inquiry
into the possible scope and depth of the tasks he designed for a certain class of
theoretical machines when devising his theory of computability in Turing (1936).
He considered digital computers one possible material incarnation of those theoretical machines – and human computers another, which served as the blueprint
for the former.
3
Turing’s theory of computability was driven by a genuinely meta-mathematical
interest: if, within the confines of a logical calculus, there is an unequivocal, welldefined and finite, hence at least in principle executable procedure for deciding on
the provability of a proposition that has been stated in that calculus, this procedure
should be translatable into arithmetical forms. These, in turn, could be broken
down into a set of simple mathematical routines that would be executable for a
human ‘computer’, that is a person with basic mathematical skills who is provided
with a set of input numbers and a set of instructions and then ordered to calculate
the result (and pass on that result as input for further computation). These inputs
and instructions, Turing’s argument continues, would be simple and unequivocal
enough to be handled by a suitably designed machine, too.
Although representing only an expressly restricted subset of human intellectual abilities, these tasks were well beyond the scope of traditional, cog-beltand-pulley machines. Margaret Boden (2006, 168f ) notices that the thought that
machines could possibly think was not even a ‘heresy’ up to the early twentieth
century, as that claim would have been all but incomprehensible. The reason is
that intellectual capabilities were thought of as essentially tied to the (most likely)
human organism whereas the capabilities of machines were mostly thought of
as being restricted to the exertion and harnessing of physical forces. More precisely, according to machine-age definitions of machines, “the primary function
of a mechanical device can be either the modification of motion (direction) or the
modification of motion and force (amplification and reduction)”, where the former
would be a mechanism and the latter a machine proper (Mitcham 1994, 170, in his
reconstruction of definitions of machines that go back to the mid-nineteenth century). A more reductive definition confines the activities of machines to “changing
the direction of motion” of matter, because “moving matter is all the force with
which machines deal” (“What is a Machine?” 1872, 39). A more differentiated