Well, here I go. Many of you aren't going to give me a lot of credibility due to me being a new member, however, I want to discuss a thought that I planning on developing. I got this wacky (or not so wacky) idea after watching Ghost in the Shell S.A.C. What if we had an A.I. that was ran on the standard computer that was able to take voice commands and use that to communicate with the user. Not only this, but it would have a virtual body that could be fully designed with top of the market 3D developing software such as 3D Studio Max or Maya. I know that I may be looking into something that has already come out commercially but I can't find anything close to how I want this to run. I have been working on a Design Document (I know that this is more targeted toward game design) due to the structure of the AI being developed, I think of this more of a tool for the advanced computer user. Most of these users are "gamers" that spend more time on average then the typical user. What do you think, I know that there isn't an A.I. coding template and most of this would be trial and error, but if there is any suggestions on what I should add to the A.I. or how I should setup installation or the graphical interface, post a comment on the forum. Anything can help, but please think about what you are saying before posting. I'm really thinking about developing this.

Why do "advanced" users want something to automate this stuff? (and last time I checked most gamers aren't "Advanced", yesterday I was having trouble telling several of them how to open a file with Notepad)

How do you plan to get this "standard computer that was able to take voice commands and use that to communicate with the user"? As far as I know, there is no such a thing at the moment. Seeing as chat bots have, for the most part, failed miserably using just text chat, I don't expect much out of a voice-controlled one.

ut2004 has bots which take voice commands and talk back to the player (in a very limited fashion). Are you trying to replicate a real human in a computer game? what is different about your idea? Elaborate please.

"Not only this, but it would have a virtual body that could be fully designed with top of the market 3D developing software such as 3D Studio Max or Maya"

don't all games do that?? We are long past 2d limits and text adventures.

For your information, do a search on AI voice, and you will get some reliable software (for the most part). Don't be so quick to give a chewing either, this is an idea that I want to build off of. Gamers just tend to know more about computers then lets say, someone that works on MS Word day in day out. Most gamers are wanting their computers to run quicker than the next, it is a computer that runs a few packets quicker then the other that determines a kill or death. Just an FYI. Voice can be implimented very well if done correctly. You know, don't you think it would be "cool" or "techy" to have an AI voice greet you when you turn the computer on? Or would you like to have your computer RAM and CPU usage said to you during a game to prevent keystrokes. You know, I'm just looking at the idea of being away from the keyboard and being just as efficient as if I was on the computer.

All you have to do is get an audio program that records voice (acid pro) and program the AI to take that recording and match it with what is said. Yeah, it may be out of the league right now, but as far as I'm concerned, without someone attempting, you'll never know!

And for the several gamers having trouble openning a file with Notepad, if you said it the same way that you replied to me, I wonder why they didn't understand. Ezbez, I'll call you out right now, if you know so much about AI chat bots, maybe you would like to respond to why they fail miserably? Instead of the B.S. that you sent previously, let's make some head way on this.

Your call.

Now my idea. This A.I. would have a virtual body. Some movies have this. Actually, Halo is another game that I got the idea from. The virtual body would be entirely 3D. This body would encompass a link to the software (such as toggles for voice commands that can be turned off or on). Not only would it be limited to this, but the user can program macros (like predefined keystrokes in a sense) where you can record the word that you want to say, and give it an action. I don't know how reliable it would be.

A good example would be:
"Player" and the AI could open the default media player.
or "Time", and the AI could announce the current time.

The AI body, could have random actions throughout the running process of the computer. Ex. The AI could "Walk off the screen" and ten minutes later, could walk back on from the other side of the desktop. Or rip open the desktop image and zip it back together. It wouldn't be a game, it would be a desktop companion.

The desktop companion, hence for laughs, Cheers, could be called by saying the AI name and he would walk back on the screen. Let's just say, right-clicking on Cheers would give the user the option to select a different skin (much like that of WinAmp) or the option to import a character that they have made to help them visualize what the user would image the character to look like. Really, all the user would need is a Mic and all the software could be put into an installation setup.
Want more information?

I have another question, what do you use your computer for?

[Edited by - Akira52803 on May 27, 2007 5:14:02 AM]

Most of us use their computer to program and / or play games. I really don't need someone to greet me when my computer starts and I definitely do not need a software that would waste my CPU/RAM trying to do things I can do myself.

Considering voice recognition, getting the system to understand single words has been done numerous times, but the real problem remains in merging these words in a sentence and understanding the meaning of this sentence. This brings us to the same problems chat bots are facing.

As for having an AI giving you tips on what to do ("Or would you like to have your computer RAM and CPU usage said to you during a game to prevent keystrokes."), that's pretty useless. Most games will bring your CPU to 100% usage and RAM has little to do with your in game actions. The AI would have to _KNOW_ what a certain keystroke will cause in game (if it was relevant...) and for that the AI would have to be aware of the state the program is. I see two options there... read the memory of the current application and try to find what's going on (almost useless, but programs like Softice were useful for cheaters back in the days...) or have every developers use a standard API to let them notify your "bot" what's going on in the game and what would each action branch to...

In other words... not gonna happen.

I suggest you try to work on smaller projects...

Good luck!

Akira52803,

I'm going to comment without much experience or authority here. I have actually thought about a similar idea. But nothing to do with gaming. Simply developing a voice recognition/response system for whatever purpose it may serve. Whether its a game feature, OS (accessibility) module, or just something cool shouldn't really matter. I find the problem itself interesting.

The thing with speech is that its not an easily formalized problem. When someone speaks, the person on the other side doesn't just hear for the communication to succeed, they need to understand as well. This is where you hit the gray zone when it comes to AI, I think.

If you want something to respond to preprogrammed sequences of words, I think thats a different problem altogether. In fact, you can find technology like this everywhere. Voice recognizing hotlines, for example. However, in the case where you want a human type of response (if thats what you're getting at) you need a lot more. Two things that immediately come to mind that seem insurmountable are:

1. You need a way of understanding what words mean. You might be able to do this for individual words, but how for combined meanings of words? In a sentence for example. Even if you construct an AI that somehow can extract the meaning from a sentence, it needs to now produce its own set of words with meaning for a response.

2. If you get past the first problem, and somehow produce a sequence of words for an answer, you are faced with the problem of making your response sound human. Splicing recordings of individual words sounds like anything but human. Although, I have read about research in this area with a computer controlled voicebox. I don't know if it was a success or failure though.

In any case, I think this problem in its essence has to do with artificial cognition -- an area which isn't very well understood. I don't want to sound negative though. Research is being done all the time, so who knows what the future will bring.

if you ever got into home automation you would be familiar with HAL(home automated living)2000(yea like the movie). it dosn't have EVERYTHING that your talking about(there is no cute little figure on the screen).

i've used HAL in a number of home automation projects in the past. its a very nice program, responds to a voice command(think star trek 'computer') or to a key press. It can be set up to give a TTS reply or use a wav file.

this program is ideal for home automation because its designed to interface with X10 devices. but it can also access any files on your computer, a minute or two to set up the voice command the the automated reply and you can have it do just about anything.

The one major difference i can see between HAL and what your talking about is that there is not AI in HAL and i dont see the point in haveing AI in such a program.

now that i'm thinking about it, the microsoft agent could take voice commands to run programs, read you e-mail, yada yada yada.

you can chec out HAL here http://www.homeautomatedliving.com/default.htm

Hey

I've looked into it before too. It depends just how far you want to go with it as if you want your computer to "understand" what you say it takes alot of work. Depends how indepth you want to go... but here is a small intro if you want to look further:

NLP - Overview

Context free grammar - grammar understanding

N-Grams - ngrams, potentually a big part of a intelligent system

N-Grams slides - ngrams again

MyCyberTwin - online chatter bot, give it some time, might turn out to be pretty cool in a year or 2

Book: Jurafsky, D. S. (2000). Speech and Language Processing. Prentice-Hall, New Jersey. - One of the better beginner book on NLP and Speech processing

You can have thousands or even millions of people continualy improving and updating an expert system or database through normal usage of query and command technology, and this can even be done transparently.

Thus I don't think 'contextual understanding' is really an issue here. As long as cause and effect are rational (the appearance of contextual understanding) then you can have that cake.

Appriciate the help, I was curious if anything had been done before that somewhat falls along the lines of what I am planning. I am attending ITT Tech for game design and I figured that this could incorporate what I am learning for a final project.

xEricx, it's not really for gaming, it's just something that I figured could be potentially helpful. Think about it, if you own your own computer and are on it every day, you know where everything is. What if you could limit the use of a mouse to that of just gaming or manual operations. The point that I am trying to get at is this, for example, you are across the room, working on something else that prevents you from going to you computer, and you want to navigate windows through voice commands, would that logically be possible?

From what I have gathered, it would be. Don't get me wrong, it's not something that would be a piece of cake, nor is that what I want. I want a long term project. Really, an algorithm based on the targeted functions and using that algorithm to plan the code, technically, all you would need to come up with the voice command and the function for that command. Also you would need to have recording software to allow for the user to create his/her own voice commands.

The AI part of it was to make something like a gif but 3 dimensional that could possible walk around on a desktop. This was more of a way to make the development more complicated. By adding the AI (and possibly responses from the AI) you could basically have a non-stop worker for windows. It's just a theory.