Which neural net has the best hallucinations?

[Read the post]


How similar are two different runs through the same network? There are probably pieces of the algorithm that “look” in random places, right?

1 Like

GoogleNet’s had the best ending, measured by my personal and unarguable yardstick of inverse caninoid pareidolia. We know androids dream of electric dogs, but can we please have the universe melt into other things for a change?

Obligatory Penny-arcade:


That’s what this guy is going to do. The reason there are so many dogs is because everyone’s running their images through the canned dog model Google already made. If instead you create a new model trained on boobs or… or… um… Okay, let’s say boobs, then any image fed through that model will turn into boobs instead of dogs.

Of course you could also do both, but that would be weird looking.


There are many paths one may choose, but they each reveal it’s doglizards on doglizards, all the way down.

1 Like

I thought that the neural net commonly known as @GilbertWham had the best hallucinations.

They weren’t always enjoyable. Stay away from ketamine, that’s my advice.


This topic was automatically closed after 5 days. New replies are no longer allowed.