Ok, so ever since I learned about neural networks and genetic algorithms, I've wanted to try all this stuff out for myself. But alas, I am but a meek artist and don't know if I have the capacity to learn to code. I know there's a lot of open source stuff out there but it's still way over my head. I just want a simulator where I can make some simple creatures that I can watch grow and teach themselves how to do various things. I can't be the only one, right?
Geddy135 Do it in 2D, it is easier than learning 3D graphics. You can work with flash+action script, pygame, ascii in the console, etc. There are genetic algorythm packages like DEEP, though I think you can easily implement it yourself. Each creature is an array of random numbers that represent different stuff, like limbs or movements. You put that in a function that calculates the new positions and gives it a score according with your criteria. What you see in the video is just a visual representation.
Think is amazing still, and when compared with the Elon Musk- sponsored Open AI 2d walker, and Deep Minds AI, that look like dead zombies, it should be at least as famous. We make a game called Kontrol that has characters that simulated legged walking over rough terrain and hills and it took years to develop. This genetic learning technique can open the way for Generalized AI, and AI that can solve more than just one class of useful problems. Seems that survival in a physical world, with and against other species and teams of them gives meaning to intelligence. The creatures exist because they competed, cooporated, loved and hated, survived and bred, they may emerge to ask for the same "meaning of life" as we do.
The program creates a bunch of randomly designed creatures from blocks and hinges. It also randomly coordinates the creatures movement with a neural network. The creature that behaves the strongest is the base design for the next round of creatures, all slightly different. This process is redone for hundreds or thousands of generations, and the result is a creature designed by evolution, not us.
Joey Ortiz "The creature that behaves the strongest" What exactly means to behave the strongest? What would the criteria be? Just watched it until the end, I'd guess that the first to hit the cube is the strongest? Cheers
this crab like creature passes the cube and perfers to beat up its opponent. made me laugh
Made me worried
"Perhaps it is mean to move the goal away just as it arrives..."
Now I feel bad.
Back around 2015-17 this video had an outsized effect on my choice of going for a bachelros in computer science. I am now working on my masters :D
I wish informative youtube videos were like this nowadays
This is both cute and deeply terrifying.
1994!!!???
Adam Hughes yeah why
Didn't know people were doing such advanced things that, long
genetic algorithm just came from a biological study and implemented for computers that can run it very fast. nothing advance about that lol
so this not ai?
@@ogvibe11 this is called "Machine Learning" so yes AI I suppose. The intelligence is derived from machine. AI = Machine Learning
Incredible
"One learned to push his opponent away"
more like punch the poor thing in the head
This was made the year before I was born.
Absolutely fascinating.
This is awesome
amazing
why do I find this funny? xD
Hokey with appropriate gestures😂
More: karlsims.com/ en.wikipedia.org/wiki/Karl_Sims
Am i the only one who is both really amused and oddly horrified by these creatures?
Ok, so ever since I learned about neural networks and genetic algorithms, I've wanted to try all this stuff out for myself. But alas, I am but a meek artist and don't know if I have the capacity to learn to code. I know there's a lot of open source stuff out there but it's still way over my head. I just want a simulator where I can make some simple creatures that I can watch grow and teach themselves how to do various things. I can't be the only one, right?
Same... I want to have a simulator like this.
Geddy135 Do it in 2D, it is easier than learning 3D graphics. You can work with flash+action script, pygame, ascii in the console, etc. There are genetic algorythm packages like DEEP, though I think you can easily implement it yourself.
Each creature is an array of random numbers that represent different stuff, like limbs or movements. You put that in a function that calculates the new positions and gives it a score according with your criteria.
What you see in the video is just a visual representation.
www.openprocessing.org/sketch/377698
Paperspace.com has tutorials you can learn from scratch .
Coding train!!!!
Think is amazing still, and when compared with the Elon Musk- sponsored Open AI 2d walker, and Deep Minds AI, that look like dead zombies, it should be at least as famous. We make a game called Kontrol that has characters that simulated legged walking over rough terrain and hills and it took years to develop. This genetic learning technique can open the way for Generalized AI, and AI that can solve more than just one class of useful problems. Seems that survival in a physical world, with and against other species and teams of them gives meaning to intelligence. The creatures exist because they competed, cooporated, loved and hated, survived and bred, they may emerge to ask for the same "meaning of life" as we do.
EVOLVE MORE. PT2
we are living in a simulation
Wow
explain what is happening...
The program creates a bunch of randomly designed creatures from blocks and hinges. It also randomly coordinates the creatures movement with a neural network. The creature that behaves the strongest is the base design for the next round of creatures, all slightly different. This process is redone for hundreds or thousands of generations, and the result is a creature designed by evolution, not us.
Joey Ortiz
"The creature that behaves the strongest"
What exactly means to behave the strongest? What would the criteria be?
Just watched it until the end, I'd guess that the first to hit the cube is the strongest?
Cheers
Salomão Rodrigues karlsims.com/papers/alife94.pdf
Bloo Bloo
Thanks for that! :)
the criteria are given by the programmer i think. maybe indirectly..but yea
what software is used...
+Zhenhao Zhou software that is no longer available :(
+John Doe What's the name of it ?
I think is AnimatLab.
This was software written in 1994 and run on a supercomputer...
A modern tablet computer might be faster than the computer this software ran on.