9 years ago
5
Google sets up feedback loop in its image recognition neural network, creating hallucinatory images that veer from beautiful to terrifying
Yes, Androids do dream of electric sheep.
Continue Reading
http://www.theguardian.com
Additional Contributions:
-
+3Submitted 9 years ago by AdelleChattre
-
+3Submitted 9 years ago by nxthesky
Join the Discussion
Some really cool images here. I found it particularly funny the mention of how the software automatically inserted arms into the dumbbells image because it didn't know that arms aren't attached to them. I don't know much about AI / machine Learning but the dumbbell situation makes me think about the difference between AI getting smarter as opposed to us being able to program more info/algorithms.
This is really fascinating.
The mistakes it makes are actually kind of adorable. I suppose there may come a time when adorable becomes terrifying.
I wonder if something similar happens inside the brain of a person on hallucinatory drugs. On a mild dose is the brain trying to find things that aren't there and is that why if you try to focus on one thing you can begin to see it. Or so I've heard.
Scientists actually believe that our brain naturally look for patterns that aren't there (even when we aren't on drugs). The arguments for this are kind of interesting (that it originated as a defense mechanism that has helped us survive in the wild), although I am a bit skeptical about that. It's not too far of a stretch to think that hallucinogenics would make it easier for our brains to pick out patterns.
I hope this takes off as a legit art form. This is such fun to look at!