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 a game called rain world, the devs were experimenting on an AI they created that was a lizard enemy with a long sticky tongue. after they tried breaking all its legs off, the lizard stayed motionless for a minute before figuring out it can use its sticky tongue as a grappling hook, which was completely unintentional on the developers part
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
Some days I think Skynet is here…some days…I think we are going to be okay.
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).” Good Lord.
Yesterday I overheard someone talking about how he was taking classes at the University of Maryland because they offer free tuition if you’re over 60.
My brain IMMEDIATELY began scripting a screwball comedy in which a broke millennial who desperately want to finish his long-abandoned degree but is drowning in student debt pretends to be a senior citizen in order to attend college for free.
I’m picturing someone Channing Tatumesque, applying age makeup every morning before he heads off to class. It’s sort of a cross between 21 Jump Street and Mrs. Doubtfire. He keeps forgetting which hip is supposed to be his bad one. His classmates laugh every time he uses slang. There’s definitely a scene where he attends a college party and busts it up on the dance floor.
He catches the eye of a fellow returning student, a woman in her 50s, but she thinks he’s like 70 and she’s already buried one husband, you know? She’s not interested in doing that again. When his charade unravels (hilariously) at the end of the movie, though, she finds out he’s actually like 30 and has abs you could bounce a quarter off. And he’s still super into her. And really, maybe it’s time she gave May-December romance a chance.
Okay so to refine this concept a little:
Our Hero is stuck in a job where he keep seeing people get promoted past him because they have a 4-year degree and he doesn’t. He can’t afford to go back to school until he finishes paying off his student loans for the degree he’s one semester from completing. If he got the promotion he wants he could pay them off a lot quicker. But he can’t get the promotion without the degree.
Along comes a clerical error in his almost-alma mater’s records which lists his birth year as 1948 instead of 1984. He gets a call from them about their “free tuition for seniors” program. “Wow, that sounds amazing!” he says. “I’ll be sure to tell my, uh, grandpa, as soon as he gets home.”
It’s one semester. If he can keep up the charade, he’ll have the degree, get the promotion, pay off the student loans. Hell, if they figure it out after the fact and come after him for the tuition, he’ll be able to afford it by then. He just needs to pass as a 70-year-old until graduation. How hard could it be?
His best friend (Chris Evans? don’t judge me) thinks this plan is insane. Our Hero is going to get caught. But Best Friend does have a little sister who’s a professional makeup artist, and she does know how to do age makeup.
She also thinks this plan is insane– people at Alma Mater aren’t stupid! Someone will notice! But she’s game to try, and it’ll be great practice for her.
She tells her aunt about this insane plan at a family dinner a few weeks later. “I mean, it’ll never work,” she says. “But I kind of want to see him try.”
“Uh-huh,” says her aunt, only half listening. She’s distracted for a reason, though– she’s going back to school, her first big independent decision since her husband died. Classes start soon! She’s excited, but also nervous.
Wackiness, obviously, ensues.
(at the end of the movie: “Am I technically your uncle now?” Our Hero asks.
“NO,” both siblings say in unison. “She’s our aunt by marriage! That’s not how this works!”)
(also, someone in the notes suggested “Senior Year” for a title, which is PERFECT.)
(other outstanding suggestions from the notes: Harry Shum Jr. as Our Hero and Salma Hayak as his love interest.)
Some time ago I made a post about internet movie host Count Gore de Vol, and it lead me to another internet horror program called Monster Madhouse (site), except this one puts special emphasis on kaiju and giant monsters. The main host goes by the name Karlos Borloff.
Above is an episode showcasing Gappa: The Triphibian Monster, that among other things also shows footage of meeting suit actors Haruo Nakajima, Bin Furuya and Tsutomu Kitagawa (also in this video).
They even have original kaiju as hosts sometimes:
And even made a silly short with them:
And I’m just scratching the surface as I’m finding out about this. Their youtube channel has hours of old fashioned public tv creature feature goodness.
Side note: These don’t have motors. They’re completely momentum/wind-powered and literally just wander around beaches unsupervised like giant abstract monsters.
Monsters and Stuff 