You might have some art history under your belt. Now experimental artists are giving the machines a chance to do the same.

It’s called Google Arts & Culture Experiments, and it takes a new angle on machine learning. The concept: let those algorithms find new ways of venturing through the history of art and human culture.

This isn’t just about the machines, either. Continuing the Chrome Experiments series, the search giant is enlisting artists and creative coders to try an inventive take on what this might mean. After all, while the machine learning may be for the AI, it’s the humans who decide how to apply it.

And if the Renaissance had perspective and paint innovations and the de Medici family as patrons, now we’ve got algorithms as medium and Google as the patron. There are some names followers of creative coding will immediately recognize, too.

Just a few examples:

You get Mario Klingemann (Quasimondo, of Germany), seeing how machine learning might interpolate between artworks from different eras and continents.

Cyril Diagne turns oceans of thumbnails into flurries and mountains of images.


There’s a TED talk, too:

The twist is, these aren’t just random arrays of images. True to Google’s trade, big data is transformed into patterns. Watching those algorithms try to “see” these images can feel a bit like watching a child first learn to draw – in a fascinating way.

I find it especially appealing, having watched the Metropolitan Museum of Art’s efforts to digitize its holdings in New York City. Art history curation involves nothing if not data.

And if lately the reflections on big data have been mostly bleak – laments about fake news and echo chambers – here we can return to more uplifting applications. This is a chance to see in a new way, not reinforce old ways of imagining.

Blog post:

  • Polite Society

    That is so cool. I just lost an hour or so trawling through the site. Thanks for the heads up.

  • Antti Tiihonen

    Google’s recently launched AI experiments have some super cool audio things that definitely deserve to be covered by CDM as well. 🙂

    Infinite Drum Machine, which presents a pretty nice way to organize and explore huge sets of audio samples:
    Giorgio Cam. Take a picture and then AI produces rhymes based on what it sees and sings it on top of a Giorgio Moroder track:
    AI Duet. Jam with an AI:

  • Daniel Tibet

    In other words, Google swallows the entire history of art into its AI hive mind, like it previously swallowed all the world’s books, and will exploit it to add value to its own products, like it does with all the world’s data.

  • Emmett

    Essentially, automating the process described by George Kubler’s “The Shape of Time”.

    I’ll agree that these sorts of ventures probably are the future of looking at the past, but I’ll also agree with the sentiment Mr. Tibet expressed above. Academia was built around the idea that knowledge should be accumulated for it’s own sake. If the research is being directed by a corporation, then the underlying drive is that these discoveries should have some underlying monetary value. That these discoveries must assert their value within a pre-established system, the hegemony that resulted in the institution that is funding the venture.

    We can’t have a radical new understanding of our past without the willingness to abandon our worn out imaginations of the future.