mobeamer: March 2010

Monday, March 29

Code: Simple 2D Javascript Map

A friend of mine asked me what is the quickest, simplest way to create a 2d map. Think top down, grid like map.

Here is 35 lines of javascript code that will do just that. Enjoy...


<html>
<style>
.tile {
background:#BBB;
border:1px solid #999;
position:absolute;
top:0px;left:0px;
width:5px;height:5px;
overflow:hidden
}
</style>
<script language=javascript>

var originX = 0;
var originY = 0;
var tileSize = 32;

function buildMap(width, height)
{
 for(x=0;x<width;x++)
  for(y=0;y<height;y++)
  {
   var d = document.createElement("DIV");
   d.className="tile";
   d.setAttribute("ID","tile_" + x + "_" + y);
   d.style.top = originY + (y*tileSize);
   d.style.left = originX + (x*tileSize);
   d.style.width = tileSize;
   d.style.height = tileSize;
   document.getElementById("divMap").appendChild(d);
  }
}


</script>
<body onload="buildMap(10,10)">

 <div id="divMap"></div>

</body>
</html>

Friday, March 19

Artificial Intelligence - Dreaming and Learning

So last but most importantly, what does the Ai do when its sensors are detecting nothing?

It has two options.

Use Active/Aggressive sensors
The first is to turn on some active sensors. We can assume that some sensor it has can “roam” or find new views of it’s environment. Let’s say we have a web parsing sensor, it reads HTML and brings back links. This sensor would not be active unless there is nothing else going on. The Ai could decide to activate it. There could be an array of these type sensors.

Dream
The second option that it has is to dream. When dreaming the Ai will:

Search it’s Long term memory for a random data variable
Use this data and feed it to a all Evaluators except for “the best”.
Any evaluator that understood the data gets a chance to process it
Any evaluator that understood the data would get passed additional data elements that are “similar” to the original”

Doing this allows the Brain to evaluate if it’s truly has the best evaluator and sensor collaborated.

Artificial Intelligence - Memory Management Example

So, let’s image that its time to bring the Ai Online. We’ve built all the components and we’re ready to go.

Here’s how the first 1,000 interactions with if the Ai was only attached to a “Textbox” sensor.

Definitely take a look at the picture as it conveys a lot more but briefly:

“Hi” comes through the textbox sensor, the Ai stumbles through Evaluators till it hits the Language one, which say “I understand this data”, the Ai says ok and lets the evaluator do its thing.

The next time the textbox sensor fires off, the Ai gives it to the only Evaluator it knows Language, language handles things smoothly. This takes place 999 times.

However on interaction 1,000 along comes a math request “3+9”. The language evaluator says…Ummm, no I can’t do anything with this. Again the Ai stumbles around until it hits the Math evaluator, who says, yeah…I can do something with it.

Now when textbox fires off, the brain will randomly ask Math…Do you understand this stuff…If Math says yes, the Brain lets it handle the text. If the Math says “no”, brain turns to Language and passes stuff off.

Summary
This is not the most efficient way and if you notice it never actually pattern matches against what comes in on the textbox sensor.

However, if a different sensor is setup for math functions this approach would lead to an efficient Ai. Math coming across a speech/text sensor would be like being asked does this spell BLUE. Humans can do it, its just…odd, thus we are more likely to answer incorrectly.

Artificial Intelligence - Memory Management

Let’s talk Memory,

A Brain is going to have to recall things, store numbers, understand relationships and all sorts of stuff that deal with “recall”. This is where the memory module will come into play.

Memory will consists of 4 pieces. All parts of the memory are available to the Brain and the Evaluators. Memory is not available to sensors and or motors.

Short Term Memory
Short term memory is for swift, context based data. It will consist of a Sensor ID, the data, a timestamp, and number of times accessed counter. At first this will be a straight copy of the sensor stack. However it will change over time. The number of items/rows in short term memory will be dictated by the amount of time it takes to search it and retrieve a specific piece of data. For example, if a query against short term memory takes longer than 100 milliseconds, the memory handler will move the 10 least accessed memories to long term memory.

Long Term Memory
Long term memory will look just like short term memory. Sensor ID, data and timestamp for when it came in. However Long term memory will have a different requirement, when searching long term memory takes more than 10,000 milliseconds, delete 10 of the least access memories.

Note: This means that the Ai’s memory will only be as good as it’s architecture. The faster the searching the more memory that the Ai can have in it’s short term memory.

Action Memory
Action memory records the chain of events that occurred from the data coming in to the motor response. It’s purpose is to help the Ai choose which evaluator to send sensor data to.

Every time the Brain sends data to an Evaluator it will check to see if the sensor and evaluator exists in the Action memory:
If it does not the Brain will insert a new row
If it does, the Brain will increment the number of attempts

When an Evaluator receives data, during its processing it will:

Increment the understood column for all rows where the same sensor and evaluator are specified
If a row does not exist for the motor used, the evaluator will insert a new row with that motor in it.

The success rate is simply a calculation of attempts/understood.

Technical Note: The Action Memory table is not normalized (its flattened), this is intentional as the inserts will be few a frequent while the seeks or queries will be much greater. Thus this table needs to support quicker seeks. I will attempt to explain further later.
Memory Data
For storing variable or relationships or even more complex “thoughts” there is data memory. It’s just a key, data paring. Which seems too simple. However, I think it would be the responsibility of the Evaluator to make sense of the “dirty” nature of the memory. I must admit that there is most likely a better solution. However keep in mind that this memory bank has to server a variety of functions and Evaluators.

Summary
So all of this gives us the memory and storage of data. I would expect that storing this and accessing this would take up about 90% of the storage and processing capacity of our Ai brain.

Artificial Intelligence - The Brain

So this picture paints the whole picture, except for the most important part, which is in the next post. But let’s make sure we have it all covered.

Here’s how it works

When the “brain” is ready to start a new tas
It will remove the top item from the sensor stack
It will determine the best sensor (this is the unexplained important part)
The Evaluator will send commands to the motor stack
It will also send memory objects and requests to Memory Handler

This continues till there is nothing to sense. We will talk about what the brain does when there is no data on the sensor, but not right now.

So, we have a little gray area, but the picture is getting clearer right?

Let’s continue…

Friday, March 12

Artificial Intelligence - Evaluators

So we discussed how the Ai will sense the world and we have discussed how it will interact with the world. How will it bridge the gap and make intelligent interactions based on what it has sensed?

Before we can cover that you have to understand that the Ai Brain is actually made up of a number of parts. Some folks would consider each of these parts as Agents or AI themselves. I will leave that debate to others that are wiser than me.

For my purposes I call them Evaluators. The Ai Brain consists of many of them. Each Evaluator is responsible for performing the following:

It takes data in, the data will always be in a rawest form
It attempts to understand this data

It will understand the data if it knows an motor action that can be given performed based on the input
Or it knows a memory object that needs to get created based on this action
Or it knows another evaluator that will understand this data

If it knows a motor action to perform based on the input, it will place the motor action on the motor stack
If it knows a memory object that needs to get created it will create the memory object*
If it knows another evaluator that will understand this data it will pass this “suggestion” along. **

Note that a sensor stack item may be sent to multiple evaluators. Some evaluators will be better at some things then others and usually there is a “best” evaluator for each type of data on the sensor stack. We will talk about this more later.

Example evaluators:

Language evaluator - responsible for holding conversations
Math Evaluator – responsible for performing calculations
Image Evaluator – responsible for identifying an image
Knowledge Evaluator – responsible for identifying relationships between objects

I would imagine that at some point the Ai brain may create its own evaluators by combining or altering existing evaluators. We will talk on this further as well.

* See memory creation post (tbd)
** See brain evaluator selection post (tbd)

Artificial Intelligence - Motors

So, we defined how the Ai will monitor and sense the world, how will it interact with it?

This happens almost in the reverse of what happens with sensors.

This time we start with the Motor Stack. The Ai places some data and a motor ID on the stack. As it “thinks” about additional task it keeps adding stuff to the stack.

Motors will continually do the same thing.

Look on the stack for an action with their identifier.
Take the data and perform their action.
Repeat.

What can a motor do? Anything. The simplest of motors may be responsible for taking text and putting it in a textbox for a user to see. A more complex motor may take an filename and display the contents of the file to a user. An even more complex motor may take a formula and graph it over time, sort of like your old T1 calculator.

Wednesday, March 10

Artificial Intelligence - Sensors

Sensors

So without senses we, humans, would not be able to function. Much less be considered “intelligent”.

In the same fashion our Ai needs a way to sense changes in it’s environment.

Think of a sensor, as a program that sits outside of the Ai brain. They monitor the environment and then pass data from those changes to the brain. So, the Ai will have a number of these sensors that will monitor the environment. T

How it works:

Step 1: Some change takes place within the environment of the Ai. For Example: A text box has “Hi” typed into it.

Step 2: The sensor detects the change and gathers the data. For Example: “Hi” is saved as data

Step 3: The sensor sends the data into the Ai. This is where the data crosses into the Ai’s brain.

Step 4: The Ai, takes the raw data and the sensor ID and puts in on the bottom of the sensor stack. The sensor stack is a “list”/stack of data which the Ai has done nothing with yet.

Some Points to Ponder:

Sensor ID: Each sensor must have a unique identifier. This is essential for the Ai as it will learn to do different things with this sensor’s data. Now just because the sensor ID is constant does not mean the sensor has to watch the same point in the environment. In the above example, the sensor could go from textbox to textbox taking in the data.

Sensor Stack: The sensor stack has all data points from the sensors which the Ai has not processed yet. This is critical as the Ai may let data stack up till it all make sense to it.

The next post will discuss how the Ai will use the stack to decide on actions.

Tuesday, March 9

Artificial Intelligence - Take One

The next few posts will be about a proposal for an Artificial Intelligence (Ai) system. It’s based on many things I’ve read around the internet as well as my own general ideas on how such a system needs to be made.

My goal is to detail the system in a few articles and then attempt to build it. If given time, a thing I have precious little of these days.

This shows the basic concept of the AI. As you can see it’s made up of 4 components. We will talk about each component in depth but from a high level you have:

Sensors – Track and record changes in the environment. This would be the point at which data enters the Ai’s consciousness. An Example: A sensor may be responsible for tracking a single text box on a given web page.

The Brain – This is where all the interesting stuff takes place, stay tuned. An Example: The brain may decide that in response to text entered it should display a picture.

Motors – The brain will issue simple commands which the motors will execute on. Motors will perform actions on the environment. An Example: A motor may be responsible for displaying a picture to a webpage.

Memory – Pieces of data and how the brain acted upon this data are stored in memory. Later on I will go further in depth with what will happen here. An Example: The word “Hi” might be stored in memory as well as the fact that the brain responded “Hi” back to it.

Some things to keep in mind:
This is an all-purpose form of intelligence, it’s not what many would consider a knowledge based system, or built to be superb in a specific category. In fact, I predict that this brain will not be superb at anything, but rather much like the majority of humans, it will be sufficient at being adequately productive.

My goal is to build an Ai which can function in the same capacity as an average human. Think of the Turing test but including tasks rather than chat.

My measure of success: If the program can return/perform in such a fashion that I can’t tell the difference between it and a human, I will have succeeded.

mobeamer

Monday, March 29

Code: Simple 2D Javascript Map

Friday, March 19

Artificial Intelligence - Dreaming and Learning

Artificial Intelligence - Memory Management Example

Artificial Intelligence - Memory Management

Artificial Intelligence - The Brain

Friday, March 12

Artificial Intelligence - Evaluators

Artificial Intelligence - Motors

Wednesday, March 10

Artificial Intelligence - Sensors

Tuesday, March 9

Artificial Intelligence - Take One

Followers

The Lab - Free Code

Favorite Posts

Posts by Category

Past Posts

Game Dev Blogs

Blogs I roll With

Must Read/Visit

Future Reading