Intelligent Agents should be good at predicting the future

2010-03-02T00:19:00.000-08:00

When thinking about Intelligence one of the attributes often talked about is ability to predict future events. I think there has been some serious amount of work put into trying to create a "reality predicting machine". I'm not sure which path people have chosen to follow to meet this goal, but if I had to wanted to create a predicting machine I would not start off with logic.

From my understanding, logic is emergent from something else. So when trying to build Intelligent or Predictive agents a hard coded Logic language does not need to be created. Correct cause & effect procedures can bee created from a neural net, which is not logical but habitual. Most successful link continue to be successful and unsuccessful links die out, creating habitual operations for the agent, which may appear logical & intelligent to an observer, when in fact they are not.

AI researchers should focus on creating habit forming agents. perhaps after the agents habitually learn how to speak they may call these habits... feelings... or un-logically caused behavior.

Big Picture

2010-01-21T08:25:00.000-08:00

When building a model of complex phenomena it's a good idea to try and find a way to represent several phenomena using one or a few mechanisms. For example: if we didn't know anything about color, we would have to create a conceptual place holder for each color with no connection between them. This list would be huge, and hard to work with. However with more research it is possible to represent all colors (even those that are between the colors in the list) using a commonality between all colors, light waves. The model of colors becomes simple, input a light wave and you get a color. In order to represent all the colors you no longer need a list, you just need 1 equation (with defined range). All colors are a variation of 1 thing, light waves.

Much of my life I have tried to find a commonality between everything. And I do mean everything: humans, animals, plants, rocks, planets, ideas, language, relationships, knowledge, science, .......

The commonality I've found between everything is: change.

This may seem odd, "change" is not a physical thing, sometimes its hard to quantify or even qualify, it only exists in our minds. [If our thoughts can imagine everything real, but reality cannot be everything we can imagine then logically it is possible to find a commonality between everything in reality through our abstract thoughts]

I am still puzzled if initial origin, or an irreducible unit is shared by everything. The Initial State query is analogous to "What started the Big Bang?", can you have something out of nothing? "Is there an irreducible unit of stuff?" might seem important when we are trying to figure out what is everything made of... However I propose that those questions are not important enough to stop progress. I don't have proof if there is an initial origin of reality or if there is an irreducible unit in reality, but i do have an example that shows that we don't need either of those 2 things for something to exist. Numbers! There is no beginning to numbers, they range from negative infinity to positive infinity. There is also no irreducible unit, no matter how many digits there are after a decimal point anyone can add another one.

So what is so great about Change? Change is the difference between two or more entities.

Change can change - The amount something has changed can be quantified, this quantity does not have to be the same. Our height changes all the time, but the amount and direction of that height change changes as well.

Change is relative - we don't need to know all of the details about an object, we just need to know how its different from another object

Multilevel Change - The amount something has changed can be treated as an entity. Basically, in terms of physics, if you have object(s) that are moving, Change exists at several levels. An agent can focus on velocity, acceleration, jerk, ....

All these "attributes" of Change are actually different views of the same property. If one of those attributes exists all of theme exist.

So what does all this have to do with AGI? - For an AGI to function in reality it needs to handle Change. Handling of everything else will emerge from this. Modeling reality is all about figuring out how things change. In other words an AGI agent will need to change its own model of reality by adapting (changing) it to actual reality.

In the Beginning...

2010-01-20T11:51:00.000-08:00

In this blog i will try to lay out my model of reality, i will try to go top -> bottom, general -> specific, large scope -> small scope.

This blog is an attempt for me to get my thoughts in order and perhaps learn something new and useful on the way.

Lets see how it goes...

My Artificial General Intelligence Research

Intelligent Agents should be good at predicting the future

Big Picture

In the Beginning...