Quite a few members have been writing about and coding for AI driven insects, amoebas, etc. and discussing using classical weight driven neural nets for their brains
I wrote this little side project as an exercise in machine learning to demonstrate both emergent behaviour and the transference of intelligence/ skills through both shown example and environment driven behaviour.
If you where to take a human baby at birth and keep it confined with no human contact would it grow up to be average intelligent human citizen? Although DNA gives us the machinery to become intelligent, we are each defined mentally/ morally by what we learn from our parents/ teachers/ peers.
Introducing B-BotYou can teach the B-Bot to do anything within the scope of its environment. To hug/ follow the outline of the walls, patrol a certain area in set patterns, avoid the blue and hunt the green or vice versa, B-Bot can learn any methods or actions you choose to teach it.
At the start the B-Bot knows absolutely nothing, no memories or intelligence of any kind.
The B-Bot is entirely sensory driven; it can see in a forward facing arc, it has a memory and the ability to move. I’ve given it the equivalent of an eye saccade just to help with the training times.
Some emergent behaviour might see if trained long enough lol…
Stubborn belief system and free will… lol (it can be a pain to teach)
Follows food then jumps on it.
Object avoidance – will run and swerve around objects to get to food.
Appears to actively hunt food – will wait for it to emerge then grab it.
Territory – the bot can self define a route or territory to patrol
Tactics – Just watch your bot and let me know what you see… anthropomorphism.
The B-Bot learns from how you play the game. I suggest you first spend 5 minutes guiding the B-Bot around the arena, turning at the walls and getting B-Bot used to its environment. There are no boundaries set in the app, the bot will leave the arena unless you show it how to turn at the walls. If it does manage to flee… it will re-spawn in the center and carry on.
Then hit the green tickbox and start teaching B-Bot to eat the green dots/ food. Try to guide the bot so it hits the food head on, as the bot learns you will notice it taking over, becoming more confident and jumping/ guiding its self toward the food. This usually starts to happen when the ‘Mem’ number reaches 2500 ish. The more you train the bot the better it gets. When the bot hits the wall, guide/ help it by using the arrow keys, teach it what to do in this situation.
The schema uses intelligence plasticity so you can always re-train your bot out of a particular trait it’s learned.
You guide/ show the B-Bot what to do by using the left/ right arrow keys for steering and the UP key for moving forward.
Once your bot is trained click the ‘move’ tick box to start the food moving, then train some more… train the bot to catch moving food.
You can then active blue food… the bot will stop when ever it sees it because it doesn’t recognise it or know what to do… train the bot to avoid blue food but eat green food… or to eat both.
If the bot stops and a question mark flashes on the left, just use the arrow keys to guide it out of a situation or towards the food.
The App is written for the Windows API and you should hopefully be able to download it from this link on my G-Drive. It’s all been virus checked and don’t worry I’m not going to take over your computer and make it part of my huge AI bot net… lol.
https://drive.google.com/open?id=1OhY1mX4xEJI0Wy1fAflaU6gCHFu3rhNWThe more time you invest teaching your B-Bot the more intelligent it will become. Lets see what you can teach your B-Bot to achieve.
At the start or after a reset always press the forward key a few times first, this just teaches B-Bot to move forward when it sees nothing ahead.
Have fun…
ED: I wrote the app and instructions in a few hours between jobs, sorry for roughness, I will be revising the above text lol.
Q: Considering the many complex behaviours you can teach B-Bot, what kind of neural nets do you think I'm implementing?
Poll