After a long ten weeks, I’ve finished the Stanford University online Machine Learning course! It’s been a solid, if mostly science/engineering orientated, introduction to designing and implementing the basic algorithms. While it usefully touched on several image processing applications (road driving, OCR, pedestrian/face recognition), in all cases there was (by necessity) a clear result that each application was trying to achieve.
To apply ML in an art context, past the ‘deep dreaming‘ quirkiness (that successfully communicated something tangible about how such algorithms operate, although I wasn’t able to fully comprehend what it was doing until I’d learnt how to code such processes) is probably going to take me much longer to conceive as one must skirt the ‘what is the purpose of art’ type questions, which is probably not the kind of thing that can be usefully defined in the way that a ML algorithm could be devised for.
For instance, I’ve seen various projects that take visual styles from one artist (Van Gogh seems to be a favourite) and attempt to derive or apply them to new or other works by different artists. This kind of thing is certainly fun, but is shallow (aesthetically, if not technically) and quickly becomes repetitive.
However, I do think there is scope for application in highly personalised interactive artworks that will either provide a deeply transformative human experience, or possibly nothing at all, which might be the most one (as an artist) can hope to achieve anyway!
If anyone is interested in exploring/discussing ideas along these lines, then do drop me a message, or please share to people you know who might be involved in such things…