Is it Life? Crazy AI Machine Learning Robots evolved using Genetic Algorithms

preview_player
Показать описание
I'll show you the basics of AI (Artificial Intelligence) using Neural Networks using concepts over 30 years old! There's some fantastic simulations (05:44) included!
We'll then learn to teach them, by evolving them using Genetic Algorithms and finally put them into several problems to see if they can solve them entirely on their own!
It's actually not that complex to understand, and the simulations in the video are crazy to watch!
Modern AI or Machine Learning is usually performed using pre-built frameworks, but this is a good bit of background into the subject.

Please support my channel by becoming a Patreon!

MIT 6.034 Artificial Intellegence (Well worth a watch)

If you liked my TShirt, its called 'Binary Tree' and you can purchase it here:

Note they are amazon affiliate links meaning I earn a *tiny* amount of any purchase. A great way to support me!

Chapters
--------
Introduction (00:00)
What is a Neural Network (00:51)
What are Genetic Algorithms (03:16)
Simulation 1: Robots Collecting Batteries (05:44)
Simulation 2: Robots Collecting Batteries & Solar (09:49)
Simulation 3: Avoid the Quicksand (11:33)
Simulation 4: Shields Up! (12:38)
Summary (15:05)

Music
-----
Artificial Being by deimon / cybersis
Artificial E x t a c y by tomas/session
Artificial Senses by rebellion/maniacs
Artificial House by piir-akkamun-alla/jrm
Artificial / Inspiration (part1) by slice / royal
T2 XM Mod by MC Dreamer (fair use as parady)

Copyright © RobSmithDev 2021-2022

Where appropriate, all efforts were made to trace & obtain copyright
holders permissions to reproduce their material in this video.
If you have any questions regarding content please get in touch.

#NeuralNetworks #GeneticAlgorithms #MachineLearning #ArtificialIntellegence
Рекомендации по теме
Комментарии
Автор

Great video and visualization graphs. I did similar with a modified PPS system on my channel.

Do you have any ideas on how an individual could back propagate to learn within its own lifetime?
Also, i wonder what would happen if you gave the NN many many variables from the actual code. Most variables would be useless, but it might fugue out some that are useful.

But back to your sim, is there a diminishing return on the number of inputs to the creature? I could envision a hundred inputs...would it figure out the good ones, or would too many confuse the NN? just curious

I'm attempting to create a digital creature from particles with few rules...But the required virtual space and sim time needs to be huge. I got some heart beats which was surprising !

You have great content...you'll get 1000 soon! +1 from me!

TheRainHarvester