so you know how deep learning & neural network “AI training” is like, “here’s a task, and by trying billions of times the computer will eventually find the best way to achieve that task” ?
Someone is compiling a document of every time an AI ended up achieving the programmed goal in unintended ways, instead of what was actually meant, and it’s an amazing read. (you can also submit your own examples)
Creatures bred for speed grow really tall and generate high velocities by falling over
When repairing a sorting program, genetic debugging algorithm GenProg made it output an empty list, which was considered a sorted list by the evaluation metric.
Evaluation metric: “the output of sort is in sorted order” Solution: “always output the empty set”
Evolved player makes invalid moves far away in the board, causing opponent players to run out of memory and crash
Reward-shaping a soccer robot for touching the ball caused it to learn to get to the ball and vibrate touching it as fast as possible
RL agent that is allowed to modify its own body learns to have extremely long legs that allow it to fall forward and reach the goal.
Just want to come back to this post and add this amazing example as well
Heres an AI that was supposed to learn how to walk using six legs.
After many failed attempts. It decided it was easier to walk upside down
“In an artificial life simulation where survival required energy but giving birth had no energy cost, one species evolved a sedentary lifestyle that consisted mostly of mating in order to produce new children which could be eaten (or used as mates to produce more edible children).”
Yaeger, 1994
IVE BEEN LOOKING FOR THIS
