Not so crazy AI?

Deliberately wrong?

I was preparing for a talk to students a while back (really must develop the pictures I took) and I asked DALL-E (an AI powered image generator) to make me a notice with the message “Don’t Panic” written on it. Above you can see the closest that I got. During the session I said that I thought this proved that “proper” AI is some distance away. But I’ve thought about it since and decided that actually AI might be ahead of us on this one.

Perhaps the creators of DALL-E don’t want anyone to be able to create fake notices and signs and so they have the system do a bit of scrambling to the letters before they are rendered.

Interesting times.