How can I make an Ai (Artificial Intelligence) program that learns what I tell it?

Question

I have a Mac and a dell so scripts or whatever are not a problem. I have a little bit of JavaScript knowledge but a lot of batch scripting knowledge. I don't need to start out with anything too large so even a cmd program would be great. I just need to figure out how I can make my program (my bot) learn.

Answer 1

Hi John



Don't start with writing your own - look on soureforge for pre-written things that you can download and install for free to see how they work - learn from them and then if you like, try to develop your own.



AI is a big field, so I would really advise to have a look at existing apps first.

Answer 2

Read my answer here:

https://answersrip.com/question/index?qid=20130508092506AAMSE72



To expand on what I previously wrote, if you want to make a program more intelligent, then you have to add more responses to user input. In the example program that I gave, the key to a better AI is more pre-programmed responses to the questions. However, I didn't stress the importance of pattern recognition. The example program just uses simple single word recognition as a means of determining a response, but you could do more. For example, you could allow the user to input whole sentences and analyze the whole sentence by picking out more than 1 keyword.



Let's talk about patterns. Life is a pattern. Just about everything in the world is a pattern because we humans generally can't deal with chaos. When you're looking at a cloud, you begin looking for a pattern stored in memory. When you're walking down the street and see the people walking around you, you're looking for patterns -that's how you recognise people. Data compression code works by looking for patterns and better compression can be achieved with more agressive pattern searching. So, if you wan't a program to be more intelligent, you have to write code that can process and handle more patterns.



As far as AI intelligence that mimics human thinking, the most important part would be to pick the brain of a psychiatrist and other professionals who study human behaviour. Those people will be able to give you the common human responses for a given situation and thus, you are armed with the patterns for your program to process information and respond in a human-like way. I'll give you an example of this. In the '80's there were some programs (e.g. Mindview for the IBM PC) which asked the user questions like how outgoing they were on a scale of 1 to 5. At the end of the questionnaire, the computer spits out (in plain English sentences) the personality of the person -which was pretty darn accurate. The program itself was very simple; Essentially the program just took the input, stored the data, and then the program spit out the proper pre-determined sentences later, based on the input. So, the psychiatrist who gave the programmer the information on how to respond to the questionnaire actually did all the work.



As mentioned before, most people have no idea what a computer is, and that's the greatest reason why all that AI research is nonsense. I don't even want to talk about the insanely bloated, slow computer code...

Answer 3

How To Program Artificial Intelligence

Answer 4

Do research on genetic algorithms, algorithms development, machine learning, Turing completeness, and probabilistic programming, Bayesian Logic.

But first, please read something about strong vs. weak AI.



There's only a limited number of ways you can simulate intelligence with such a small amount of resources.

You can collect and parse data and evaluate information based on the information you collect.

You can record certain actions, their frequencies, and crunch numbers based on what you collect.

You can even have your algorithms optimise and alter themselves over time.



But you're using a computer.

You are doing math, only; until someone is able to successfully describe intelligence in terms of mathematics, you will not be able to write an actual learning computer. That's not to say you can't write a really convincing weak AI, but it's important you realize what you're trying to do before you go for it.



On a tangent, personally I think that the farther away we are from this the better, an actually intelligent computer raises important moral questions about the nature of the object we have formed. Is an intelligent computer not just as worthy of life as we are?

Interesting questions, if you like philosophy.

Best of luck.