My theory on how to create AI was not pursued as an actual project because I was missing vision recognition. I say that at the top of that page. So I put my project on pause since 2002. I am reopening the project now that researchers at the Austrailian Centre for Visual Technologies(acvt) have created a software package that converts video into 3d models called VideoTrace.
The rudimentary AI that I am describing is simply an imagination space for the computer. In order to do most thinking one must picture the environment in a 3d space. Without a 3d imagination, nothing further can be built on top. My previous theories explained why this is the case, and I'm not going to rehash it. If you don't understand why imagination space is needed for artificial intelligence, you're not alone. Many people don't get it. If everyone knew this was needed to make AI, I wouldn't be writing this webpage. In short, the rudimentary AI that I am making here will be a building block for creating more advanced AI on top of it. Just like 3 ghz computers weren't the first ones we had, we're not going to have SCIFI AI off the bat... but it will rapidly follow because this rudimentary foundation AI is the hard part. And rest assured, the hard parts to do this rudimentary AI are already done. All I need to do is piece together some software parts and it will be done. For the rest of this HTML page, the AI I am building isn't SCIFI AI, but instead a foundational AI which is just an imagination space.
In case you're wondering what an imagination space is, I can run you through a quick explanation. If I tell you to think of a basketball, you probably are picturing a basketball in your mind. The background is irrelevant, and the make of the basketball doesn't matter much either. In short, you have an idea of what a basketball looks like. In your mind you have a 3d model of a basketball, and that is how you can imagine it. You've seen a basketball before, and that is how you know how to imagine it in your brain. Essentially this software will do the same thing. I'll use VideoTrace to capture a basketball model into the computer. I can then display the basketball on all black background. Finally I'll have to write code that can take in video and identify objects so the AI knows when it sees a basketball.
So I need the software to:
A) Take in 3d models and store them in a database(a bit of work, but VideoTrace claims it can do it)
B) Use video to take in reality to the computer(trivial, I've already done this before)
C) Run software to identify 3d models from the video(difficult, but I think I can hack it)
D) Create a scene based on all the 3d models it sees
If I can accomplish the above things, the foundation of AI is complete.
You may be wondering how exactly I can claim that the lofty goal of AI is complete in just those simple steps. I'll give a quick explainaton. Lets say you do step A and digitize everything in your house including the walls, floor and ceiling. Next you should be able to boot up the AI that does steps B,C, and D. Next you take a camera and walk it around your house so that the computer is able to digitize everything in your house and now has it in its imagination space. Do you see that you have the foundation of AI now?
Think about it this way: Your AI now has exact dimentions and descriptions of everything in your house. If you made a robot that had some basic motor skills, you could tell it to go from one room to another(not in natural language yet, just in code). Then the robot would navigate itself around obsticles in your house, and navigate from one room to another. The robot was able to navigate using a built on camera and the AI software.
The end all be all revelation of this AI foundation is that it is all stationary nouns and maybe adjectives for version 1. Once the AI can understand stationary nouns in a 3d imagination space, we can build foward to teaching it verbs. I'm getting ahead of myself here, but once the AI understands some verbs, you can move ahead to a natural language interface. And once the AI has a natural language interface, it can then soon learn from books. The coolest thing we'll see first about AI that understands natural language is that it will be able to turn story books into video that you can watch. Another cool thing it could do is that people could use it to program via a natural language interface. Anyway, I'm getting ahead of myself. There are limitless uses for Artificial Intelligence.
The key here is for me to let you know that a 3d imagination space is needed for AI to be realized. If you think about it, it makes sense too. We've never had the technology to identify nouns through video before and create a 3d imagination space. So it makes sense that we've never had the technology to make AI. If you'd like to help create FOSSAI, I'll be making a SourceForge.Net project for it as soon as they get back to me on the status. I'm also waiting on my copy of VideoTrace(I'll bother them once a week until they send me a copy. LOL). So if you want to help, email me at: FOSSAImail @ yahoo.com
