Machine learning and copyright: we still don’t have the big picture; much less the final word
The impact of the widespread popularity of increasingly sophisticated machine learning tools able to generate images, texts or videos based on a simple written request is beginning to be felt among artists, writers and other creators, particularly over the issue of intellectual rights.
On the one hand, the current doctrine with respect to intellectual property leaves little room for discussion: the US Copyright Office says that the product of an algorithm is not eligible for intellectual property protection because it is not a human creation, in the same way that animals have been refused such protection by the courts. In short, says the US Copyright Office, intellectual property protection implies that the creation originates in the intellect of the person claiming it, and that machines or animals cannot, therefore, do so.
That would automatically make the creations of software such as DALL·E, Midjourney, Lensa AI, Stable Diffusion and others devoid of intellectual property rights, therefore leaving them in the public domain. That said, there are nuances. Many artists claim, for example, that their works are being used to feed the huge repositories of tagged information that algorithms use for their creations, and they are not being compensated. As anybody who…
