We have already talked on several occasions about systems capable of creating images from text. We see it daily with Dall-E, with imgcreator, with Stable Diffusion, and with many other platforms that we’ve been showcasing here at WWWhatsnew.com over the last few months.
The operation in all is the same: we give it a text and it generates an image. The question is, would we do something similar with the same text?
There is an online game at glasp.co/dalledle/ that can get us out of doubt. It shows several texts and two images, images that have been created by DALL-E from one of the proposed texts. What we have to do is select the text that generated the image, thus obtaining a score that will be displayed on the right.
Logically we are not talking about texts like “a fish flying over a house”, something that would generate an image that is very easy to guess. The texts are abstract, nothing tangible, things like “our future is in the minds of children.”
The idea of the game is, on the one hand, to help understand how a machine “thinks” when creating images, and on the other hand, to verify if we have something cybernetic in our neurons, if we think as an algorithm of this type would .
A Wordle-inspired guessing game with images of DALL·E and inspirational quotes as a message.
In the end we are always left with the question «why did DALL·E create this image from this quote?, to answer it in the session between friends on Friday night.
