A Stable Diffusion demo generates four images from the prompt "Mickey Mouse in the style of Dalí drinking coffee in Paris during dusk with Minnie looking at him longingly." The results show dreamlike scenes with limited color palettes and occasional anatomical distortions. The speaker argues AI is a creative tool that depends on human imagination to write prompts, not a replacement for artists.
Good morning, still here at Disney. You can see like the crazy incredible views in the background here. So I tested this Stable Diffusion thing, right, which is sort of like DALL-E, but basically the idea is that you enter a prompt, a series of words, and this AI model will output a series of images. So here's the prompt I input. I'm going to read it here, so I'm going to look off Basically, it was Mickey Mouse in the style of Dalí drinking coffee in Paris during dusk with Minnie looking at him longingly. And I chose that for a couple reasons. One, obviously I'm in the Mickey mindset right now. Dalí, because he's an interesting artist that I feel a kinship to. And I also wanted to see Paris and Minnie looking at him longingly, cuz I wanted to see if AI could create in a visual medium. Mm-hmm. Both emotion as well as environment and scene. So if you click through this episode, you can see the 4 images that were output, but I'll try to describe them. Essentially, it was— the color palette wasn't huge, so I don't know, maybe 10 primary major colors were represented. A couple of the arms were off-kilter on Mickey. It did look like Mickey and Minnie. So that was pretty good. And then in one in particular, it looked more, a little more dreamlike. Um, and it did capture sort of the coffee, the Parisian background. The other ones weren't as great. Um, so for the non-sighted people out there, hopefully that description helps a little bit. For the sighted, make sure you click through. So, you know, talking to some of my artist friends, there is the conversation happening about how well does this get, does it disrupt our talent? And I don't think so. Because it still requires your creative leap to come up with that prompt. So you could argue, oh well, in the future AI will create it. So sure. I think as artists, just our paintbrushes change. And so our paintbrushes will continue to be a way that we both create models, create prompts, create whatever, in order to take that which is invisible and make it visible. All right, that's it. Have a great day and, uh, hope to see you on the internet.
