This is rather compelling, great sleuthing and testing. I wonder if you could train it to write shorter ones useful for alternative image descriptions? Like Vanessa I winder what happens when we start feeding it AI created images.... not expecting it to be a detector, but in a near future where what is AI generated and what is human generated is one big mixed bag.
I dont think it needs google image search, there plenty of others out there (TinEye, whatever Pixsy uses)... But since AI image generators are based on training of large sets of captioned images, I bet that's what is being leveraged.
Fascinating! Can't wait to start playing. Like you, I wonder how we can use this feature for learning?
It might be fun for students to try Bard on their own art.
This is rather compelling, great sleuthing and testing. I wonder if you could train it to write shorter ones useful for alternative image descriptions? Like Vanessa I winder what happens when we start feeding it AI created images.... not expecting it to be a detector, but in a near future where what is AI generated and what is human generated is one big mixed bag.
I dont think it needs google image search, there plenty of others out there (TinEye, whatever Pixsy uses)... But since AI image generators are based on training of large sets of captioned images, I bet that's what is being leveraged.
I asked it to create a map and it refused. I thought it would have activated Google Maps.
I like the idea of AI reading AI-generated images.
PS thanks for the Noah Smith image and link.
I've been reading him more often lately. I don't always agree, but I appreciate his curious mind and energetic style.
I'd be surprised (incredulous) if Bard didn't draw on Google Image search. What about submitting AI generated images?
I should do that next.