Distinction Between Methods in AI-Created Art

Hi everyone,

This article on AI-generated art reminded me of the class discussion on AIs holding copyright. 

The article differentiates between two forms of AI art. In the first, the artist feeds specific images to the AI and then choses the best images out of the many that the AI creates. This process involves the artist exercising their judgement to influence what art the AI creates, and discretion again at the stage of selecting images out of the AI’s outputs. The artist may also need to change the algorithm if they are not receiving the desired images from the AI. Author Ahmed Elgammal describes this process as ‘the artist and machine collaborating’. 

In the second method, however, Elgammal’s lab has created an algorithm (AICAN) which is programmed to create art with very little human involvement, as a ‘nearly autonomous artist’. The algorithm is fed thousands of incongruent art images and is instructed to both learn general style and reject any image it creates that is too similar to an existing image. Elgammal says he ‘has no control over what the machine will generate’ and for this reason, at exhibitions, he credits AICAN solely for each work.

This differentiation between processes makes me wonder how Canadian courts would approach copyright here. There seems to need to be a distinction between artists using AI as a tool and AI itself generating the images with very little human contribution beyond the initial algorithm. Because ‘skill and judgement’ is required for a work to attract copyright under CCH, does this mean that the images produced by AICAN are not subject to copyright? Or is the ‘skill and judgement’ required in creating the algorithm grandfathered into each image the AI produces, even without active human intervention?

Beyond the copyright issue, the images are interesting to look at!

Article: https://www.americanscientist.org/article/ai-is-blurring-the-definition-of-artist