Page 272 - The Black Box Book

would be just as important as the reference to the physical dimension
of creative work, or to the fact that creativity as a socio-cultural prac-
tice has a historical dimension. However, one thing always seems to be
certain in this context: AI does not have all this. It cannot achieve all this
and never will.
But the possibility of distinguishing between man and machine is not
based on a static or stable anthropological difference. Instead, it marks
a historical process that includes the possibility of profound, unforeseea-
ble transformations. Accordingly, projections about the future capabilities
and properties of computers are fundamentally just as questionable as
all-too-convincing determinations of what machines will never be able to
achieve. In both respects, as history shows again and again, experts have
often been proven wrong. Besides, what AI can and can`t do was already
addressed by Alan Turing and indeed most of the typical projections of
AI prominent in current discourses are already to be found in his writings
(Dotzler, 1989).
POPULAR IMAGINATIONS OF AI CREATIVITY
There is a recurring tendency in public debates to discuss AI only in
the utopian (or dystopian) mode. The computer scientist Bertram Raphael
(1971) once suggested to use AI as an umbrella term for problems “which
we do not yet know how to solve properly by computer” (p. 101). In other
words, once computers successfully master certain tasks or a solution
finally emerges, they are no long considered AI.
There are several reasons for this phenomenon. For example,
empirical research and development of AI technologies are strongly under
the spell of their cultural imaginaries. And the latter, in turn, concerns
not only the progressive potentials of technologies but, as is well known,
above all their negative implications.
How often are the real-world consequences and dangers of AI in
the near and distant future illustrated with images of HAL or
the Terminator? But whatever shapes our notions of intelligent machines,
societal fears will not be overcome by simply giving them a cute, infan-
tile appearance or enabling them to perform amusing actions. When
Boston Dynamics literally made its robots dance at the turn of the year
2020/21, this performance had obviously something uncanny about it.
The emergence of this specific feeling certainly has to do with the fact
that, on the one hand, machines resemble us increasingly in their outward
appearance, while, on the other hand, they continue to be recognizable as
machines. The latter applies for the art-project, the drawing robot Ai-Da
(2019) as well as for the robot Ava in the movie Ex Machina (2015). But
independently of this, this sense of unease is also influenced by the fact
that media culture repeatedly confronts us with imaginations in which
the difference between man and machine is not discernible. A spe-
cial case in this respect is the film The Stepford Wives (1975). While
the women ´replaced by machines´ are visually identical to their old human
272
appearance, the motivation for their replacement is precisely that they
should behave in a fundamentally different way, which is to behave like
robots in the literate sense, corresponding to the fantasies or desires of
their husbands.
More importantly, the latter scenario is no longer confined to
the spheres of art and popular fiction. 2018, with its Duplex telephone
assistant system, Google was able to demonstrate that machines can
communicate with humans within certain limits without revealing their sta-
tus as non-human beings. In view of such developments, the expectation
that after decades of failed attempts, computer systems will finally pass
the Turing test appeared again. Indeed, it would be a milestone in the
history of information technologies, but we are not there yet, although AI
technologies like GPT-3 raise expectations that this could change very
soon.
FORMS OF AI CREATIVIT Y
As a reminder: Back in 1950, Turing proposed to replace the ontological
question of whether computers can be intelligent or not with a test by
means of which it should be determined whether computers can evoke
or simulate the impression of intelligence. More than 30 years ago,
Margaret Boden conceived of machine creativity similarly (1991/2004).
The machine creativity would not be concerned with determining whether
and to what extent the machines are creative or not, but whether they
can, within certain limits, create the impression of creativity, by which she
means producing ideas or artifacts that are new, surprising, and valuable
(Boden, 1991/2004, p. 1). Accordingly, Boden differentiates three catego-
ries of creativity: combinatorial, exploratory, and transformative. These are
elaborated by Boden in relation to a “conceptual space” as a “structured
style of thought” (Boden, 1991/2004, p. 4) and discussed on the basis
of different software applications with reference to their performances
by and with computers.
As examples of programs that master a combinatorial creativity,
i.e. they can combine known things in unknown ways, she used, among
others, JAPE (Boden, 1991/2004, p. 3). This is a computer program
for generating punning riddles, the original version of which was devel-
oped by Kim Binsted in 1993 as part of her Master of Science thesis at
the Department of Artificial Intelligence at the University of Edinburgh.
The second type of so-called exploratory creativity aims at the ex-
ploration of a potential, which is to a certain extent already present within
a conceptual space but has not yet been discovered or exploited as such.
As an example, Boden mentions the composition program Experiments in
Musical Intelligence (a.k.a. as EMI or EMMY), developed by David Cope
in the 1980s. This program was already able to analyze a composition
and break it down into different components, identify the style based on
its patterns, and rearrange the different components into new patterns
without reproducing anything identically.
273