Algorithmia

Song: BenBen - Algorithmia
Video: Nate Dorr

2022 | 4 min

A machine learning-powered music video for BenBen's Algorithmia, off of the eponymous EP released in Fall 2022.

Do neural networks listen to simulated rock bands? Algorithms have dictated much of what we see and hear for years already; they've only much more recently become able to reproduces semblances of it on their own. With the advent of adversarial learning (pioneered by chess AIs trained through near-endless competition against eachother), digital simulacra have become more and more eerily convincing: songs extended through extrapolation, faces invented, and so many machine-generated avocado armchair designs. Especially over the last six months the tools for fully AI-created images (based only on generalized knowledge from data sets consisting of millions of captioned pictures) have rapidly evolved, diversified, and improved, allowing more and more generative artists to explore the form. But with algorithms already warping artistic production through the feedback loops of streaming and social media, what may come of vertically-integrating algorithmic distribution with its own ability to create? Or, in the deeper exploration of machine-generated art -- the interest of which comes as much from its mystifying errors and omissions as from its successes -- will the essential significance of the guiding human hand become only more obvious?

In the time since this was created -- in January 2022, eons ago in AI imaging -- we've seen this sort of thing become all the more pervasive, and raise significant concerns about how these models borrow from and incorporate the styles of individual artists. As the systems have improved and become more problematically precise, however, I see lingering charm in the flawed and unpredictable results we reached here, where there's no escaping the specifically algorithmic nature of the work.