ATC Revisited: Tom White
As part of BCNM’s recent transition to Zoom webinars, artist and researcher Tom White was gracious enough to take our last ATC lecture online. We couldn’t think of a better person to wrap up our series on robo-exoticism as White tuned in live from Wellington, New Zealand to discuss “Neutral Abstractions.”
White is an artist-engineer who has worked with artificial intelligence for over 25 years, and came perfectly equipped to address modern AI interpretation and abstraction. Given how COVID-19 has begun to reframe notions of physical space, BCNM faculty Ken Goldberg attests there are new ways of thinking on robots — seeing them as safer alternatives and ideal helpers in a time of crisis.
A few core themes run through White’s work: machines have their own way of seeing; we can create art for and by machines; and art allows us to see how machines perceive the world. White situates his work in an extensive art history trajectory, beginning with modernists of the 1920s. He was further inspired by predecessors such as Harold Cohen, whose generative drawing systems formalized and embodied his drawing techniques, and 1950s pop artists who popularized screen printing techniques.
His recent pieces, such as Treachery of ImageNet (2017) and Perception Engines (2018), rely on training sets to form its final images. Training sets utilize thousands of individual images to embody a single concept. These projects tested how modern AI algorithms perceived common objects like cellos, electric fans, binoculars and more. White pointed out how each of the individual pieces had the same starting point — that is, the same parameters and drawing engine — but had different end results.
White toys with the idea of meta art and agency. While he personally sees himself as the creator (using his machine learning software as a tool), there are other AI art approaches that depict the machine as an independent actor. He ultimately questions the uncanny concept of exploring the internal worlds of machines and developing artificial systems in our image, even when they do not work in the ways we expect.