I was curious, given all the current attention being paid to Dall-E and Dall-E Mini, and Craiyon, to see what any of these would come up with if I input concepts related to intercultural dialogue.
Obviously, the very first term I input was intercultural dialogue. You receive 9 images for any text, and most of them were blurred and distorted beyond much use. But I’m attaching the images both Dall-E Mini and Craiyon created. Interestingly, related terms, such as multiculturalism or intercultural competence, look much the same: a group of evidently diverse people – but never two individuals.
Whereas dialogue looks quite different – just speech bubbles. And many of the images created at my request, such as that for peacebuilding, were quite useless (nearly all of the designs for that simply showed a badly distorted vision of the UN logo, for example).
I thought others might find it interesting to see what the results looked like. Clearly, given that these AI programs simply search the internet for relevant text and images, we all need to be posting more content for it to discover that includes both descriptions and visual depictions that come closer to what we think it should find.
For those who are not yet familiar with these programs, Dall-E is a tool to turn words into images using AI, but it is in beta and not yet publicly available. Dall-E mini is the less competent but publicly available version. And that has already morphed into Craiyon. All of them work by looking to see what images are attached to words across the internet, and then editing those images into something new. Apparently, so long as you credit the source, the images are yours to use and post.
If others try any of these systems and obtain interesting results, send them along in an email, and I’ll create a later post to showcase them.
Wendy Leeds-Hurwitz, Director
Center for Intercultural Dialogue