Looking at Abstract Art with Microsoft CaptionBot, April 17-28 2016
An excerpt appears in CLOG (2018)

For Not Really Confident, I've run 267 images of abstract art through CaptionBot (© Microsoft 2015), an image caption-generating AI that came online late March 2016. In CaptionBot's "own" words, "[I] was created to showcase some of the new capabilities of Microsoft Cognitive Services. These new capabilities are the result of years of research advancements ... Specifically, I use Computer Vision and Natural Language to describe contents of images. I am still learning, so sometimes I get things wrong."

My starting place was this list of abstract painters on Wikipedia; most images of the artworks themselves were found using Google's image search. I was familiar with maybe 15% of the artists in advance, and selected images by looking at several works by each and trying to find a decent example that summed up a main body of work, or style. Images were downloaded, then uploaded to CaptionBot for processing. You can also provide CaptionBot with an image's URL, but I opted for the upload route on the off chance the AI might search the contents of related pages for context-clues. I then took screen shots of the resulting captioned-images and organized all the before / afters into some shared folders.

As the process went on, I thought of some related works (poems, exhibition strategies, etc) which you can see under the Ideas section above. If you have any corrections, questions, or ideas shoot me an email.

Contents: