This document will be the closing capture statement of the Concept Development Phase. Requires customer approval.
I was tasked with solving the NLP problem back in 2003. I came up with a solution that was only 86% accurate. To increase the accuracy I suggested using machine learning with a human element. This was not a good solution do to the human involvement. They wanted a solution that solved the problem on its own. I pondered this problem for over a year. I came to the conclusion that the computer had to understand the words that it was reading to make sense of the document. So I asked the question “How does one understand words?” I realized that we understand words based on experiences. Then I asked “How does one gain experiences?” Through interaction. Well, how does one interact? One interacts through inputs and outputs. So you sense the world around you. You then react to the world to achieve a certain result. This result is based on desire. Finally, I guess all state the idea. This problem of NLP turned into a solution to solving AI and that's the proposal. Create True Artificial Intelligence.
Create artificial life.
There is not target market. The goal is just to achieve something no one else can.
It will be able to learn, react, and retain memory.
Scientists.
This is an ideal system. 3 42u racks loaded with clearcube 8 blades setup with dual quad core xeons. There will be 13 clearcube blade servers per rack. There will be a network switch and a kvm per rack. There will be 2 42u racks with hp storage works maxed out on storage and a 1u server with 4 hdds to control and add to the storage space. Each 42u will contain 13 1u and hp storage works configurations. This will also have a kvm and network switch.
The project should take 5 years if funded. The budget for the total time will be $2.1 million. Within the first year, the hardware will be aquired and the based code for the system should be completed. The next year will consist of trouble shooting the issues in the system. The next two years (year 3 and 4) will be spent training the system. Year five is a precaution to ensure the system is stable and unable to dominate human kind.
Most scientists belief that the human mind can be solved through a simple mathmatical solution. They all focus one heuristic(neural nets, rule engines, etc). My solution is a combination of solutions.
Skills required are people that understand the human mind beyond question. Also, they will need to understand the current state of AI, this includes neural nets, belief nets, Dempster Schafer nets, rules engines, cluster algorithms, genetic algorithms, data mining, and other heuristics.
This solution may not work. Also, I may never get funding which will hinder the development.
The goal, to create Artificial Intelligence. It's feasable and I believe that this design can create AI.