When you’ve got artificial intelligence (AI) that craft poetry, illustrate children’s books and turn your selfies into Renaissance paintings, it’s pretty hard to stop messing with it and marvel at what it can do.

Now kick that up a notch: There’s an AI that can generate images you won’t even think exist. Like a “baby daikon radish in a tutu walking a dog,” “cube made of toothpaste” or “stained glass window with an image of a white eggplant.” Yeah, nothing turned up on Google, but this AI can create a ton of images like that.

Developed by OpenAI, DALL·E is an AI trained to generate a set of images using a text prompt with specific words or phrases that users can switch around. It’s a transformer language tool, designed to translate text into images made from scratch.

“We find that DALL·E is able to create plausible images for a great variety of sentences that explore the compositional structure of language,” reads the description on OpenAI’s website.

The images aren’t always accurate, but often close enough to the prompt. You can instruct DALL·E to alter the shape, texture and color or specific everyday objects, like this tetrahedron of bubble wrap.

It can also draw multiple objects on a single image, recreate busts of well-known figures, generate cross sections, apply 3D render and even conceptualize action movie posters from the 1920s to the distant future.

As oddly amazing as the results are, you can’t enter your own prompts yet. DALL·E takes instructions from pre-existing prompts and sets of words, but you can have fun messing around with it and see what weird images it can create.

But AI like DALL·E aren’t just made to amuse us with weird photos. The developers also plan to use models like DALL·E to dissect how AI can relate to societal issues such as the “economic impact on certain work processes and professions, the potential for bias in the model outputs, and the longer term ethical challenges implied by this technology.” This could mean not only the future of creative pursuits but also the broadening of our understanding and use of promising technology like AI.

You can try it out for yourself here.

Photos from OpenAI