I'm making an A.I. program! Does it sound promising?

Well, let me start off by saying, I'm fairly new to coding, in fact I have been learning java for about 2 months now. I have taken it upon myself to make an A.I. chat bot thing, as I saw no others that functioned in this way. How does it function, you ask? Well, you input a sentence, it deconstructs it, figuring out what part of the sentence is what, (subject, verb, participle, whatever) then determines it's reaction to it. Not it's response, it's REACTION.

So, let's say, for instance, you said "Hello, I hate you, please go die in a hole," it would look at it and think"Ok, they just said hello, that is a greeting, that is average, and changes nothing (except it will maybe make me say a greeting as well). Wait, what's this? they just said 'I hate you', 'I' means they are talking about themselves, 'hate' is a 'bad word/emotion', and is used towards 'you', which is actually me, so they are expressing a 'bad word/emotion' towards me." The "please go die in a hole" part would follow a similar train of logic. "Please" is generally kind and or polite, "go die" is a verb, and dying is a bad thing. "In a hole" has no other context, so it is being used with the verb phrase "go die", (it's more complicated than this, but I don't want to make this post 50,000,000 characters) so they are asking the program to do something bad that it is not already doing in a place where it is not, meaning the person is probably insulting them, and is not using please in a kind/polite way.

The program would then construct a sentence based on what it knows, and what the human just said, and print out that sentence for the human to reply to. I do not know how a lot of "artificial intelligences" work, but I doubt it's like this. I am really exited to start working on this, and I would like to know what you think :D

-Thanks, GoodName

It does sound promising. One of the key features of AI is the ability to "learn" and modify behavior accordingly. You could probably implement that quite easily by attaching "score" values to words and phrases and using those values to compute an overall response value. For example, a word or phrase with positive connotations would have a positive numerical score, while a word or phrase with negative connotations would have a negative numerical score. The analysis would apply some function to the the scores to arrive at the response value, whose sign and absolute value would determine whether the program interpreted the user input as neutral (0), slightly friendly (>0), very friendly (>100), very hostile (< -100) etc.

Then to implement the learning, you "tweak" the scores or the function each time the program runs and ask the user whether the program's response was more realistic or less realistic. You then use that user response to decide whether you want to keep the old scores or take the new, "tweaked" ones. Over time, the program should appear to become "smarter".

@goodname: C++ would probably be a better choice for a powerful application than java. It is a C-based language like java, so it should not require as much time for you to learn the basics of it as java did.

Anyway, that is not really learning anything. There is no prediction, either. Your program idea is one which evaluates a sentence by its components then selects a response based on what it already knows. If that is its final version, that trait would serve to further illustrate that it is not learning or predicting anything. Additionally, the metric you are utilizing for intelligence is that of a human perspective (good vs bad. i.e. moral meanings to words). Synthetic intelligence does not necessarily need to pass the Turing test to qualify as intelligent in every way. Keep in mind that I have not seen your code, so it could be your description that needs clarifying.

It is good that you are learning about artificial intelligence as there is a shortage of experts in that field. If you are willing, I suggest you study machine learning.

Stanford University offers a free course on this subject on Coursera.

https://www.coursera.org/learn/machine-learning/home/info

Make sure it has this: https://www.youtube.com/watch?v=dI5PLZ7wmS0

Here's what I think, and here is what exactly is going to happen. AI programming is extremely difficult. You're going to give up and get nowhere. This reminds me of people who set out to learn a language to make games. They start learning a language, then realize that programming is hard, and that game programming is even harder, so they get discouraged and give up.

In the unlikely even that you actually do keep going, defying all of the odds against you, see Practical AI Programming in Java. It's a free book, under Creative Commons.

@samusbeliskner: Anyone can learn.

@Stesilaus That is a very good idea :D I was definitely planning on using user responses to tweak it and get it to sound more human, but the final version would not still be learning, at least, that's my plan. You see, I want it to function like a human having a conversation. Inside of ever conversation it has, it would lean stuff about the human talking, and everything the human talks about. It would also start getting sad, bored, impatient, and start trusting the human less, if they behaved "poorly", and the same goes for happiness, interest, etc. But it would not use that information in the next conversation it has. You see, I don't want this to be like cleverbot, who just draws from some 60,000,000 conversations it has already had, (not dissing cleverbot in any way the people who made him are ridiculously smart) I want the code to be it's brain, and this is the only way I can think of where it can learn IN THE MOMENT, and keep that information for use in the rest of the conversation.

@BranKetra I think I'm going to stick with Java for this. For what I'm trying to do, I think it's plenty powerful. In answer to your concerns about it not learning or predicting, I honestly think that having the program utilize only itself rather than some massive library is a good way to mimic human conversation. Again, in its pre-finished phases, I will be testing it with everyone I know, and hopefully getting feedback from them to change the program for the better. And you see, it doesn't just select a response, like "Hello, my name is bob", it constructs it, modifying and crossing words based on all of the different variables. This actually allows it to be versatile and work its way through almost any situation. (well, that's the hope when it's finished) Oh and thank you for the link, I'll look into that :D

@SamusBeliskner I know this project is very ambitious, but I am a very ambitious person. I also think I'm very smart, (not bragging, but I honestly am) and very determined, and don't think I will abandon it. Your statement about people wanting to do something with programming, trying it, seeing it's super hard, then giving up is a very sad but true one. It was actually the opposite for me, at first I thought programming would be some insane thing that would be completely impossible to do a thing with, then when I started, I realized that if you didn't know something, you look it up, spend 10-20 minutes figuring out how to use it, and boom, you at least have a shell of the feature you want. I fully expect for this to take 1, 2, even 3 years to be even slightly human like, (hell, it'll probably be 4) but I'm prepared to keep chipping away at it till it's finished.

-Thanks, GoodName

Edit 1: @Azure64bit Wait, how the hell did the pupa get out, I disabled it right after- I've said too much.

@goodname: In that case, it is not artificial intelligence; it is a chat bot.

@BranKetra May I ask what would make it an A.I. then?

Oh, and @thegerg can you go more in depth?

-Thanks, GoodName

@goodname: Because you're programming responses.

You can't just program responses into a computer and when it does what it's programmed to do announce that it's AI. Doesn't work like that.

Still waiting for a sexbot.

But I'm NOT programming in responses. I'm giving it the capability to analyze a sentence then CONSTRUCT its own reply. I would never in anyway say "when the human says X, you say Y." It chooses what to say itself then says it. There will be NO sentences that it already knows.

-Thanks, GoodName

@goodname: As I briefly said previously, the ability to learn and predict is an indicator of machine learning. On the subject of intelligence, the Turing Test is a standard evaluation of computational system ability. Computers can hold vast amounts of data, but they can only know what they are explicitly told.

Here is some more information:

http://www.ucs.louisiana.edu/~isb9112/dept/phil341/wisai/WhatisAI.html

Ok, so are you saying if it doesn't have machine leaning, it's not A.I.? I read a lot of that link you gave me, and by the looks of it, Turing himself said that if a machine can act human, (and pass the Turing test) then they could be considered intelligent. I don't see any difference between machine learning and pre-equipping the program with whatever it would learn with machine learning. Other than machine learning being a hell of a lot quicker.

-Thanks, GoodName

@goodname Next time you want us to read something, try use paragraphs. I refuse to read a wall of text.

@bforrester420 Sorry, I'm an idiot and I was tired when I wrote that. I hope it looks better now :D

-Thanks, GoodName

Much

Depends on your general skills.

With only 2 months of experience (unless you have had experience with other programming languages), this is a very heavy task to take.

How good are you at mathematics, How good are you at computer algorithms, how good understanding do you have of neuroscience?

Of course, if you have the drive to learn, this has potential.

@magicalclick What would the test accomplish? And thank you for wishing me luck, I really appreciate it. And when it's ready, I'll be sure to send you a link :D

-Thanks, GoodName

Edit1: @magicalclick I just saw your other post now, and yeah, I don't plan on doing work that is limited to anything less than my full potential. And you know, that's the great thing about this, it's not a ridiculously complicated program, (not that it's simple) it's just going to take a ton of time to code. I think once I finish this, I'll administer the Turing test to the best of my ability, and post the results and see what people think about it. And, as a note, thank all of you for your support, you've really encouraged me to do well, and I honestly didn't know at first if this would be at all feasible, but I have no doubt now that I'll produce *something* in the least :D

It sounds interesting, but I don't know enough about the subject to know how it would work.

@magicalclick Yeah, I'll give that a try when it's ready :D

-Thanks, GoodName