Art generated by Artificial Intelligence (AI), or more accurately Machine Learning (ML) is increasingly making not only technology news, but also being reported in the mainstream mass media. While computer generated art is by no means a new phenomenon, its increased exposure throws up both opportunities and challenges in equal measure.
Nothing new under the sun
Machines and Analogue computers have been used to generate art since the 1950’s, from simple pieces such as Ben Laposky’s 1952 work ‘Oscillon 40’ (above), through to the sale at auction in 2018 of the ML algorithm code generated ‘Edmond de Belamy, from La Famille de Belamy’ (below) created by a Parisian art collective known as ‘Obvious’.
In March this year, a concert took place, whereat music composed by Bach and AI was performed. The audience were invited to signify whether they thought the music was genuine Bach, or created by a machine. Whilst they were generally correct, there were times when, to that audience at least, the difference was indistinguishable.
Is it Art or ‘AI-rt’?
These examples, amongst many, bring up legitimate questions about what constitutes Art. Definitions consistently refer to the process of creativity; though vary on whether ‘human’ creativity is an essential component. The argument could be made that even the most complex ML art-generating algorithm requires a degree of human input at several stages, from code writing, to data selection and input, to accepting / discarding results. (Not to mention human decisions about what is a valid work of art, as opposed to unrecognisable abstract nonsense.) The Ganbreeder App, used in a previous post on MLReporter.com demonstrates this conflict well, producing results at times mesmerising, ethereal and beautiful, at other times, nonsensical, ugly and even repellent.
ML / AI Resurrection
Not as fanciful as it at first sounds. Using over 6000 film frames of Salvador Dali, the eponymous museum in St Petersburg, Florida has created a series of 125 ‘deepfake’ interactive video clips of Dali, culminating with the ability for the AI Dali to ‘take a selfie’ with museum visitors. Meanwhile researchers from the Samsung AI Center in Moscow have developed a machine to generate realistic talking heads, based on specific individuals. Perhaps the most noteworthy art-related creation to date is a series of versions of the Mona Lisa, generated from one single frame, demonstrating how she might have looked moving and talking.
There remain more questions and considerations as AI and ML are deployed in fields of artistic endeavour. Alongside debates about whether this is Art or not, are questions such as authorship, ownership and copyright. As ever technology proceeds apace, while perceived required regulation lags behind.