Mind blown -- A.I. generating images from a verbal description
Phil Pilgrim (PhiPi)
Posts: 23,514
It's rare that an advance in technology totally awes me, but this one does:
Anyway, the article describes an A.I. program that, given a description, will create an image matching that description. The images are of photographic quality and completely original. I can't even ... This totally blows my mind.
Here's an image I scanned from the paper edition of the article that's not included in the online article. The description that produced the image was, "High-quality professional photograph of an elderly clown staring at himself in the mirror." (Enjoy it now before the copyright police catch up with me! 
) The program nailed it!
-Phil
Comments
My friend, Ryan Clarke (we met while working for Parallax), usually has a Twitch stream on Friday nights. About three weeks I joined in. He noticed and said, "Hi." A moment later he directed a person named Gary to say "Hi." Gary did, and he and I had a brief conversation. About 15 minutes later I learned the Gary is an AI. Of course, being the stream is attended by hackers, Ryan encouraged everyone to try to break Gary. There were occasional glitches in Gary's matrix, but some of the responses were just hilarious, especially for those asking Gary to write sonnets about transistors and buffer overflow attacks. Ryan had to create a Python wrapper to allow the Twitch chat box to send messages to his AI API, and he's stated he's going to try it with the graphical version at some point.
From my perspective, it seems like the text-based AI is a mechanism to be used for off-hours tech support by companies willing to train the AI with their product knowledge.