We've seen plenty of weird an wonderful optical illusions over the past few years, and with the advent of generative AI, we've started to see examples of illusions created using various generative models. Usually these take the form of objects that look a bit like other objects – but here's one of the most impressive examples of AI-assisted optical illusion design we've seen so far.
Popular science YouTuber Steve Mould recently posed a challenge to viewers, asking whether it's possible to create an image on a grid of 3x3 squares that can transform into another image when said squares are rotated (a little like a rubik's cube). It seemed impossible to do so – until AI stepped in. For more AI inspiration, take a look at coverage from our recent AI Week.
In a new video (above), Mould shares how two AI pioneers contacted him separately, each having successfully used AI to make images that look different in different orientations or different arrangements of parts. Ryan Burgert and Daniel Geng demonstrated how diffusion models can effectively deduce crossover characteristics of two images, output a single image that effectively contains the essential elements of both.
In other words, this is something that would be practically impossible to create without the help an AI, which can, as Hackaday puts it, "fill in the noise until a plausible enough image begins to appear."
And, as isn't always the case with AI-generated 'art', this application of generative models is proving a hit online. As one YouTube commenter puts it, "Okay, hear me out. THIS is AI art. Not people using AI to just generate whatever they put in a prompt. But actual human creativity and ingenuity using AI as a tool to create something which previously would have been extremely difficult, if not impossible."
