Development of
the computer program, Chesthetica, started in 2006 as part of my
doctoral
research in computational aesthetics, a subfield of artificial
intelligence (AI). It incorporated a computational aesthetics model to
evaluate aesthetics or beauty in chess move sequences in a way that
correlated positively with human assessment. This means that many
thousands
of chess problems or puzzles could be ranked automatically in terms of
typically perceived aesthetics but in a very short period compared
to humans doing the work. After completing my PhD in 2009, Chesthetica
was further developed so it could also compose original chess problems,
but this work was just preliminary and only a few were generated in
2010 (posted
online a few years later).
While
computers could already beat world chess champions, they could not
compose even functional chess problems on their own. This
piqued my interest in computational creativity which is also a subfield
of
AI. In particular, to a new approach I called, the 'Digital Synaptic
Neural Substrate' (DSNS). Named so because it uses data from seemingly
unrelated domains (e.g., photos, paintings, music, chess game
sequences) and combines them into 'strings', i.e., a standard
'substrate', that enables the transmutation and regeneration of
creative elements. This
is
analogous to how a
human brain (synaptic connections between the neurons etc.)
might get inspired from
being exposed to various creative objects as stimuli. Despite having
the
word 'neural' in it, the DSNS has nothing to do with machine or deep
learning (or artificial neural networks); this is a common
misconception.
Since
2014,
Chesthetica has been using
this approach to compose
thousands of original chess problems entirely on
its own with no limit
in sight. In effect, it is doing the equivalent of thousands of hours
of human creative labor. The aesthetics model is used optionally as a
filter whereas
the DSNS is what serves as the 'spark' of creativity. In general,
Chesthetica simulates creativity in chess as experienced within the
spectrum ranging from real game
sequences to traditional chess problems. Collectively, these composed
problems may be referred to as 'chess constructs'. This caters to a
large and
diverse audience, and covers a wide range in terms of interesting,
challenging,
and educational ideas or concepts to be found in the game; however
unusual or rare some of them may be.
It remains to be
tested, experimentally, if the DSNS could also be as effective in other
domains that could benefit from the automation or simulation of
creativity.
In 2020, a new
type of machine learning, which also does
not
use artificial neural
networks, was implemented to automatically select generated
compositions based on personal user preference. Chesthetica's
constructs can
be used by human players and even composers for training or
entertainment, and by various types of researchers in their studies.
They also likely
help keep your brain healthy and mind
sharp.
You
can find more resources on these topics
here
and the links below.