New type of game, thinking....

General Forums Off-topic Chat
1

I need help with a strange proposition, using blender to code a real brain…

I need a system, that randomly sorts through node configurations, assigning things “observed” by an Ai’ into a list, that is sorted semantically, then uses those objects to “think” doing random things until it meets a condition, I have a system using a “sensor cone eye” a "reactor(eye sees candy apply force to body)"and a “Body-stores variables and is used as parent” and a “Head-vision cones are attached, and collision designates feeding”, so it can appy real forces based on it’s motion, gather resources and learn patterns, based on meeting needs, eventually I hope to write a weighted adjusting hierarchy of needs…

Some of the nodes will be populated by actions, like attempt a cyclic motion etc, (like my lizard’s run cycle)
but the more cones you have, with different behaviors triggered and ran through, the Ai could solve a goal,

So randomly spawn creatures, or objects that use rules to meet needs,

then when a rule is “solved” a diffident list has that node configuration saved, as a semantic action and these could also be grouped by the type of need,

then an algorithm compares the saved “Node seeds” and looks for similarities, while the Ai’ is “sleeping” then data is purged, and the cycle starts again, with the new days goals/objects

So this system could eventually be used to gather object data, think about what each part could do together, or apart,
and see if it can find a solution, like touch object property “food” , catch property lizard

so it re-iterates using these “semantic random seeds” populating and connecting logic nodes, until a condition is met :slight_smile: think of a thinking boid in 3d…

Some one poke some holes in this? fill in the gaps? thought is just as important as code on this one…
Starting in the right place is better then blindly stabbing at the dark

2

someone help?
I have a hardware solution, but not a budget :slight_smile:

3

You know… I’ve been thinking about doing something like this for a while - granted, my purposes may differ from yours (video game A.I. compared to fully operational A.I. assistant).

Unfortunately, I am not skilled enough to assist you with this huge task. I JUST started developing my first A.I. (programmed with specific limited actions). But, I do have ideas that I can share with you, if that’s alright.

But first, may I ask what is the purpose of this A.I.?

4

A lot of reasons

Well first off, to solve in game simulation problems, by “watching” it’s algorithms it is making, like new path-finding algorithms etc, Second to learn about the nature of thinking, and fourth to eventually help people communicate see etc by understanding the Interface system / creativity of the brain
have you seen my lizard rig?

5

My daughter has autism, and very very smart, but can’t speak, she uses sign language(she made up most of it), and grabbing your hand and leading you to her issue of the moment, and she is bright, smiley and intelligent, but, she is 2.75 years old… and can already read like 12 words, but that is not enough to speak, if they knew more about the “seed” of the neurons output /input… it would be very valuable to humanity…

another idea, have a cuddle fish watch TV in a FMRI machine, and then study the input Vs Processor Vs skin color and shape output
and see if you can pin down anything about linkages / communication…

6

more on the subject :slight_smile:

7

Moved from “Jobs > Volunteer Work” to “Misc: > Off-topic Chat”

This thread is not an offer for work, it’s soliciting a discussion.

8

There is an open project called OpenCog. Ben Goertzels work could be of use.

9

That could well do you some help. Nice project by the way - hope you sucess.

10

The greatest issue I see here. no nodes!! :slight_smile: @ open Cog

we are nodes (neurons)

If you can create a system that generated nodes based on it’s environment. It could randomly populate the nodes with rules(pre-programmed or inferred rules)

add in semantic memory relations and a system I call memory trust level (basically how many times the memory factors as true/ worked vs attempted)

anyone know how to write on the API side?
anyone know how to use a randomess seed to populate a node tree based on semantic relations?
Help?

Repeated simulations till successful = a stubborn thinker who will not give up

by generating new rules, we can evaluate there real world use as well.,

we need Code block - the observer - intakes info about the scene and sets up logic nodes based on it. as well as generates rules based on motions (like gravity etc)

the planner - sets up logic bricks in order to achieve a goal

the executor( a real world robot) - test rules for validness using real world versions of scene items

So generate logic map and populate some nodes with -scene items- , -scene infered rules-, -pre-programmed rules- and “force patterns” pre-programmed and OTF (on the fly)
and build a semantic relation table for rules, and scene items,

test generated rules, refine, repeat with new rules, scene items etc, evolve

11

I’m not sure how relevant this is now, but Dr. Goertzel tweeted it out several months ago. It’s nodular, with nested hierarchies that filter information. It’s a little over my head so…

12

Very Cool! I think he may get a kick out of some of the behaviors I have found, from the equivalent of 6 neurons…

13

I cant understand what you are talking about !

14

Ok, So , Logic bricks can form novel behaviors, from some very simple linkages,

Imaging you have a scene, with a few objects in it, and a “AI”

it has a vision cone, and can “view” around it, it builds a list, and attempts to define rules about the objects,
Now give it a goal(like touch property water), and have it randomly build “test” scenes applying forces, moving things etc, until you solve, now save the random pattern seed, repeat, eventually, you will end up with many interesting behaviors intended to attempt to solve a goal,

I think this is how a baby works, and how we “learn to learn” is by figuring out new algorithms to generate new simulations, essentially re-working gradually the logic node pattern and the content of each block,
so we solve faster with each new problem,

15

I think you are a scientist or going to be the next Stephen Hawkings. .! :smiley:

16

I think I am a meat Popsicle(


)…

I think honestly that I am 75% Spaz 25% genius
I think everyone is probably smarter then me, but I am dumb enough to keep my gears turning…
I think about things no one else cares about,
so for the most part, not alot of people can identify with me…

17

i had an idea of developing Blender into a software that reads our minds use the best tools to shape something that we have thought…i mean a Device which reads our brain signals…finds out the image…and with the help of a Super AI builds it in Blender…! so that we no longer need to use keyboard and mouse…!