24/7 AI Livestream: Evolving Neural Networks with a Genetic Algorithm
Welcome to the Evolution Chamber, a 24/7 livestream showcasing a custom-built artificial intelligence simulation. The creatures you see on screen are not following a script; they are controlled by individual neural networks and are learning to survive and find food through a process of evolution.
HOW IT WORKS:
🧠 The Brain: Each creature is powered by a feed-forward neural network with 66 inputs, 16 hidden neurons, and 2 outputs. You can see a live visualization of the active creature's brain on the right-hand panel.
👁️ The Senses: To navigate its world, the creature relies on a sophisticated sensor suite:
A 7x7 forward-facing grid that detects food directly in front of it.
8 directional "radar" rays that sense food at a distance in all directions.
An internal "clock" (a sine wave based on its age) that helps prevent it from getting stuck in repetitive loops.
🧬 The Learning Model: The AI learns through a powerful hybrid evolutionary strategy:
Lifetime Adaptation: When a creature successfully eats food, its brain is immediately mutated—a small, random change is made to its connections. This allows it to learn and adapt during its own life.
"Best-of-the-Best" Evolution: Each creature has a limited lifespan (MAX_AGE). When it "dies," its final score is compared to the all-time best score. If it set a new record, its brain is saved as the "champion." The next generation begins as a perfect clone of that all-time best brain, ensuring that progress is never lost.
WHAT TO WATCH FOR:
Behavioral Strategy: Early generations might spin in circles or move randomly. Watch as later generations develop more deliberate and efficient "hunting" patterns.
The Live Brain: The panel on the right is a window into the AI's mind. Green lines are positive weights, red are negative. The brightness of the nodes shows their activation level. You can literally see the AI "thinking" as it processes sensor data.
The Scoreboard: The stats on the left tell the story. The "REWARD" is the current creature's score, while "BEST" is the high score it's trying to beat.
This project is an ongoing exploration of evolution, machine learning, and emergent behavior. Feel free to ask technical questions in the comments!
TECHNICAL DETAILS:
Language: Written in a BlitzMax-style language.
AI Model: Custom Node-Based Neural Network with Sigmoid activation.
Evolution Strategy: Hybrid model combining "Mutate-on-Success" (Hill Climbing) with Elitist Generational Selection.
Follow the journey! Like, subscribe, and hit the notification bell!
#ArtificialIntelligence #NeuralNetwork #Simulation #MachineLearning #GeneticAlgorithm #Evolution #Programming #Coding #24/7 #Livestream #Tech
Welcome to the Evolution Chamber, a 24/7 livestream showcasing a custom-built artificial intelligence simulation. The creatures you see on screen are not following a script; they are controlled by individual neural networks and are learning to survive and find food through a process of evolution.
HOW IT WORKS:
🧠 The Brain: Each creature is powered by a feed-forward neural network with 66 inputs, 16 hidden neurons, and 2 outputs. You can see a live visualization of the active creature’s brain on the right-hand panel.
👁️ The Senses: To navigate its world, the creature relies on a sophisticated sensor suite:
A 7×7 forward-facing grid that detects food directly in front of it.
8 directional “radar” rays that sense food at a distance in all directions.
An internal “clock” (a sine wave based on its age) that helps prevent it from getting stuck in repetitive loops.
🧬 The Learning Model: The AI learns through a powerful hybrid evolutionary strategy:
Lifetime Adaptation: When a creature successfully eats food, its brain is immediately mutated—a small, random change is made to its connections. This allows it to learn and adapt during its own life.
“Best-of-the-Best” Evolution: Each creature has a limited lifespan (MAX_AGE). When it “dies,” its final score is compared to the all-time best score. If it set a new record, its brain is saved as the “champion.” The next generation begins as a perfect clone of that all-time best brain, ensuring that progress is never lost.
WHAT TO WATCH FOR:
Behavioral Strategy: Early generations might spin in circles or move randomly. Watch as later generations develop more deliberate and efficient “hunting” patterns.
The Live Brain: The panel on the right is a window into the AI’s mind. Green lines are positive weights, red are negative. The brightness of the nodes shows their activation level. You can literally see the AI “thinking” as it processes sensor data.
The Scoreboard: The stats on the left tell the story. The “REWARD” is the current creature’s score, while “BEST” is the high score it’s trying to beat.
This project is an ongoing exploration of evolution, machine learning, and emergent behavior. Feel free to ask technical questions in the comments!
TECHNICAL DETAILS:
Language: Written in a BlitzMax-style language.
AI Model: Custom Node-Based Neural Network with Sigmoid activation.
Evolution Strategy: Hybrid model combining “Mutate-on-Success” (Hill Climbing) with Elitist Generational Selection.
Follow the journey! Like, subscribe, and hit the notification bell!
#ArtificialIntelligence #NeuralNetwork #Simulation #MachineLearning #GeneticAlgorithm #Evolution #Programming #Coding #24/7 #Livestream #Tech
