Use of speech recognition and learning algorithms in Computer games

783058842956ab1b8f495cab05f64a37
0
sujay 101 May 02, 2012 at 15:56

Hello Guys,

I have not seen games which are responsive to human speech.
It would be great to have such games.

Also learning algorithms must applied in games.
For example,If we are playing a tennis game.there should be a player who can only be played by the computer.And the more we play with the player,the player should improve.Just as we improve with each game.
will this not make the game more engaging and enjoyable??? :)

Yes ,also Speech recognition and MAchine learning are still immature disciplines.So building such games is a challenge.

8 Replies

Please log in or register to post a reply.

8676d29610e6c98d6dd2d9c38528cd9c
0
alphadog 101 May 02, 2012 at 18:35

Two reasons:

  1. It’s a harder problem than you think. Siri gets more wrong than right. Imagine with hotheads in an FPS shootout.
  2. It’s overkill in many situations. Skyrim by voice? “Swing sword”, “swing again”, “swing again”, “dammit swing again!”, “swing”, “swing”, ad nauseum. As for machine learning, here’s the true secret: you need to let the player win. In some situations, if the computer learns too well, the game becomes hard and the player gives up.

You need to make games fun first, technological marvels second.

783058842956ab1b8f495cab05f64a37
0
sujay 101 May 02, 2012 at 18:47

@alphadog

Two reasons:

  1. It’s a harder problem than you think. Siri gets more wrong than right. Imagine with hotheads in an FPS shootout.
  2. It’s overkill in many situations. Skyrim by voice? “Swing sword”, “swing again”, “swing again”, “dammit swing again!”, “swing”, “swing”, ad nauseum. As for machine learning, here’s the true secret: you need to let the player win. In some situations, if the computer learns too well, the game becomes hard and the player gives up.

You need to make games fun first, technological marvels second.

So you mean to say MAchine learning will never be part of games??
It can be incorporated.The learning can be made adaptive too..:) Like if a person is losing a lot..the game can reduce its tempo a bit..:)..and let the player win…So the key is making the temperament of the game adaptive..:) May sound ambitious ..but I feel is innovative too….:)

A8433b04cb41dd57113740b779f61acb
0
Reedbeta 167 May 02, 2012 at 18:57

Many games incorporate some form of dynamic difficulty adjustment, trying to judge whether the player is doing too well or too badly and aiming to keep them in the so-called “flow state” where they are challenged, but not too much. Do a google search for “dynamic difficulty” or “flow state” and you’ll find lots of people talking about it. That’s not really machine learning, though - just a negative-feedback control system.

783058842956ab1b8f495cab05f64a37
0
sujay 101 May 02, 2012 at 19:17

@Reedbeta

Many games incorporate some form of dynamic difficulty adjustment, trying to judge whether the player is doing too well or too badly and aiming to keep them in the so-called “flow state” where they are challenged, but not too much. Do a google search for “dynamic difficulty” or “flow state” and you’ll find lots of people talking about it. That’s not really machine learning, though - just a negative-feedback control system.

Thanks a lot.I do feel computer games need Innovation to keep going.

6837d514b487de395be51432d9cdd078
0
TheNut 179 May 02, 2012 at 20:17

@Alphadog

It’s overkill in many situations. Skyrim by voice?

*ahem* Lights please :D

I think the first step would be to build realistic agents before incorporating dynamic difficulty. Often the later is implemented in a rather basic fashion, where either the agent or the player is handicapped. I agree it would be interesting to see more work done in this field though. I think like physics where CPU resources were once limited, AI might soon find itself with access to enough memory and horsepower to perform computationally heavy algorithms that enable realistic decisions and reactions. For example, agents reacting to the direction of sounds, echos, hearing heavy breathing or other player educing sounds, visual queues like seeing shadow movement, items misplaced on the ground, etc.

783058842956ab1b8f495cab05f64a37
0
sujay 101 May 03, 2012 at 15:49

@TheNut

*ahem* Lights please :D I think the first step would be to build realistic agents before incorporating dynamic difficulty. Often the later is implemented in a rather basic fashion, where either the agent or the player is handicapped. I agree it would be interesting to see more work done in this field though. I think like physics where CPU resources were once limited, AI might soon find itself with access to enough memory and horsepower to perform computationally heavy algorithms that enable realistic decisions and reactions. For example, agents reacting to the direction of sounds, echos, hearing heavy breathing or other player educing sounds, visual queues like seeing shadow movement, items misplaced on the ground, etc.

Yes.Thats what I meant.Games need innovation.Over the years have seen great innovations and improvements in Physics and graphics.But As you mentioned gradually need to incorporate agents resposive to sound,light and many more things…:)

8676d29610e6c98d6dd2d9c38528cd9c
0
alphadog 101 May 03, 2012 at 17:42

@sujay

So you mean to say MAchine learning will never be part of games??

Who said “never”? Not me. I said that 1) it’s often not required, meaning it doesn’t add anything to the game and may detract from it, and 2) it’s not easy, relatively-speaking.
@sujay

I do feel computer games need Innovation to keep going.

You are speaking from a technophile/technologist bias. Minecraft is an immense success. Angry Birds has made buttloads of money. They’re all fun in their own way. None of them involve complex machine learning or speech recognition.

Please note, I am not saying that one cannot or should not make a game that uses machine learning and/orspeech recognition. I’m just saying that they don’t mean that games will “improve” because of their inclusion.

B5262118b588a5a420230bfbef4a2cdf
0
Stainless 151 May 04, 2012 at 11:08

Machine learning has been used in games for decades. It has a few applications that make it viable.

Genetic algorithm’s are used a lot and I suppose are a form of machine learning.

Voice recognition has been tried in games, I think the last one was “Tom Clancy’s Total War” or something like that.

Never works very well

Innovation is a word much misused in games. I cannot count the number of times you read a press release that says something like “<game> features an innovative <feature> never seen in games before” and when you actually look at it you find that all they have done is change the attack button from “A” to “X” or something else as trivial.

We do need innovation, but we aren’t going to get it through the major companies. The risk involved for them is just too great.

So it’s down to the indies to come up with the next big thing.