Using Stochastic Processes to Help Humanize Artificial Intelligence

In developing examples for my book on HTML5 Canvas (HTML5 Canvas for Dummies) I experimented with using stochastic processes to improve the realism of an animated displays. Early versions of the displays appeared rigid and artificial. Adding stochastic (random) variations brought the displays closer to real life and made them much more fun to watch.

Stochastic processes are already important to AI. Stochastic Gradient Descent combined with Backpropagation is used to iteratively modify data values passed between neural network nodes in order to minimize the error between the output of the neural network and the correct/true result.

Recently, I've explored whether stochastic techniques can be used to help humanize AI. Humans are not robots. We don't mindlessly pursue objectives without variation from a given path. A more human-like AI would certainly make the man-machine interface more pleasant to deal with than a purely robotic one. There are certainly AI applications where we don't want these kind of random variations taking place ... in autonomous vehicles for example. 

You can experiment with one of my displays that employs stochastic processes ... click here to see it in action - once it's started, click control to see the variables you can modify using your keyboard.

View fullsize

P Versus NP Complexity Theory

The P Versus NP issue deals with whether every problem whose solution can be quickly verified by a computer can also be quickly solved by a computer. This is a major unsolved problem in computer science.

In common, practical terms, it deals with how to identify problems that are either extremely difficult or impossible to solve.  In order of difficulty from easy to hard, problems are classified as P, NP, NP-Complete and NP-Hard.

So why do we care? When approaching complex problems, it's useful to have at least some idea of whether the problem is precisely solvable, or if an approximation is the best that can be accomplished. 

View fullsize

Creating a 3D Effect in 2D HTML5 Canvas Games

HTML5 Canvas doesn’t include a built-in 3D context. Without using an add-on product such as WebGL (www.khronos.org/webgl), 3D has to be simulated using the 2D context.

Some games require the sophistication of a true 3D engine like WebGL. However, many games built using the Canvas 2D context can be significantly enhanced by simulating aspects of 3D motion.

In the Ballpark sample game from HTML5 Canvas for Dummies, the thrown baseball is simulated moving away from the ball thrower toward the ball player along a simulated z axis third dimension. The z axis 3D effect is accomplished by decreasing the size of the baseball in proportion to it's movement upward along the 2D y axis.

You can play the game yourself by clicking here.​  To see the full code, right click on the game web page and select View Page Source.

To simulate 3D movement using this technique, the following steps were used:

1.    Assign a z axis variable to each new object.

Along with the xPos and yPos variables, assign a zPos variable as in code section S2 assignments for baseballs:

     ball[b].zPos = zStart;

2.    As the object moves, update the z position.

In the sample game, in code section R8, the z position is calculated based  on the y axis position (yPos) and a z sizing factor (zSizeFact):

     var newZPos  = (ball[b].yPos/canvasBL.height) * zSizeFact;

     ball[b].zPos = newZPos;  

3.    Fix the z position under specified conditions.

In the sample game, the baseball is simulated to move along the z axis only away from the ball thrower. Also, to simplify the simulation, only movement up the y axis is used to increase z axis movement. So, when the ball starts to move down on the y axis, the z position is fixed. This code is in sections R9-R12:

     // FIX Z position of zFix is greater than zero.      

     if(ball[b].zFix > 0)  {ball[b].zPos = ball[b].zFix};

     // Z FIX check for not already set.

     if(ball[b].zFix == 0)

     {   // Z FIX conditions check. 

        var newYPos  = ball[b].yPos;

        if(((newYPos > curYPos) || (ball[b].bounce)) && (curYPos < zFixMaxY))

        {  // FIX Z POSITION at current level and no smaller than minSize.

           ball[b].zFix = Math.max((curYPos/canvasBL.height)*zSizeFact, minSize);   }   }​

HTML5 Status

The World Wide Web Consortium (W3C) has announced that a final recommendation for HTML 5.0 will be released in the 4th quarter of 2014. In spite of the "recommendation" terminology that W3C uses, this is official release of HTML5. This is a big deal and great news. Browser developers will have a firm definition of HTML5 they can use to bring their products up to a complete level of support for this latest version of HTML.

W3C has also defined the follow-on version, HTML 5.1, to be released in the 4th quarter of 2016. This release will include features that have been pushed out in order to get release 5.0 completed in 2014. Without this stragegy, W3C was looking at an indefinite specification development timeframe. 

Browser developers are not waiting for 2014 to begin implementation of HTML5. You can use www.html5test.com to check the status of your browser to see which features are supported and www.caniuse.com for a summary of the status of support by all the major browsers.

What is HTML5 Canvas?

The Canvas feature of HTML5 is one of the most exciting new developments in web based capabilities in many years. Here are some key aspects:

HTML5 - Canvas is a part of the latest revision and improvement of the previous version (HTML4, released in the late 1990's) markup language for displaying web pages on the Internet.

Canvas Tag - The <canvas> tag joins the other HTML tags (there are around 100 now) as a way to include single or multiple Canvas areas on a web  page. 

Bit Map Display - The Canvas area you define with a <canvas> tag is a bit map display. You can manipulate individual pixels to draw objects, images, animations and video.

JavaScript Control - The JavaScript language is used to control what is seen on the Canvas. The JavaScript application code is placed between the <head></head> and <script></script> tags of the web page.

Browser Based - The computing needed to create a Canvas display is done within the browser. In other words, it's client based, not server based. The means that creating a Canvas application is simpler than one requiring server based programming. It also means that the computing power needed to generate the Canvas display is not concentrated in single servers. Each user's browser handles its own work.

Animation - Canvas applications can generate animations. This is accomplished using callbacks from the browser. The Canvas application designates one of its functions to be called from the browser at specified time intervals. During each of these callbacks, the application draws a new Canvas display, moving objects slightly. When viewed as a continuous flow of Canvas displays, an animation is generated.

Audio/Video - Audio and video can be incorporated into your Canvas applications. Moving objects can create sounds and video images can be melded into other application objects. This is done without the use of any audio or video plugins.

Gaming - Game developers have all to tools necessary to create compelling HTML5 Canvas games.

Fun - Putting aside the technical aspects of Canvas ... it can be described as just plain fun. An HTML5 Canvas is a place where as a developer you can express your creative energies and as a user you can have expanded and enjoyable experiences.

For a more details look at HTML5 Canvas, watch for my upcoming book: HTML5 Canvas for Dummies.