Beta

Slashdot: News for Nerds

×

Welcome to the Slashdot Beta site -- learn more here. Use the link in the footer or click here to return to the Classic version of Slashdot.

Thank you!

Before you choose to head back to the Classic look of the site, we'd appreciate it if you share your thoughts on the Beta; your feedback is what drives our ongoing development.

Beta is different and we value you taking the time to try it out. Please take a look at the changes we've made in Beta and  learn more about it. Thanks for reading, and for making the site better!

Machine Learning Allows Actors To Create Games That Understand Body Language

Unknown Lamer posted about 2 years ago | from the i-can't-let-you-do-that-dave dept.

AI 30

ptresset writes "Goldsmiths college is developing technology with natural responses to human interaction. The technology enables video games characters to move in a more natural way, responding to the player's own body language rather than mathematical rules. The hypothesis is that the actors' artistic understanding of human behavior will bring an individuality, subtlety and nuance to the character that it would be difficult to create in hand-authored models."

cancel ×

30 comments

Learned that first (1)

glebovitz (202712) | about 2 years ago | (#40980393)

Naturally

Question (2, Interesting)

2.7182 (819680) | about 2 years ago | (#40980397)

What exactly makes this require machine learning as opposed to some other method?

Re:Question (2, Insightful)

Anonymous Coward | about 2 years ago | (#40980499)

What exactly makes this require machine learning as opposed to some other method?

Hey that's so weird, the submitter had the exact same thought you did! I know, right? Crazy world, man.

It's as though he anticipated this and included the last link in the summary for the sole purpose of answering this question. It's so amazing what you can learn when you take a few seconds to read what's in front of you instead of rushing to get an early post. Cosmic!

Re:Question (3, Informative)

jxander (2605655) | about 2 years ago | (#40980501)

Two things :

1) Human learning isn't special. "Machine Learning" is a buzzword that sells clicks, or whatever metric TFA is after...

2) Humans already know what body language is natural. We might not know exactly how to express/quantify what is or isn't natural, but we sure as hell know it when we see it. Hence: uncanny valley. If we can program some basic keys and triggers into a computer system - have it learn "yeah... that's too much eye contact, you're creeping me out" - we can not only make more realistic games (i.e. by not having to hand-program every bit character in the back, but rather just have them mill about and follow the standard conventions of human interaction) but it would be a big step for real-world applications, and having our future droids not look like c3po.

2a) Seriously though, play a BioWare game. While most of their games are fun and contain varying levels of good/great writing ... the way characters voice and face sync up (or rather, don't) can be more than a bit unsettling.

Re:Question (2, Insightful)

Anonymous Coward | about 2 years ago | (#40981413)

"Machine Learning" is a buzzword that sells clicks, or whatever metric TFA is after...

Machine learning is a branch of AI. It's as much a buzzword as geometry is (which, as you know, is a branch of Mathematics, not a buzzword, although it's true it's easier to sell a machine learning paper than a geometry one). Very basically, machine learning means feeding a lot of data to an algorithm, it learns from the data and can give good enough solutions for new problems.

Re:Question (1)

Anonymous Coward | about 2 years ago | (#40982279)

Very basically, machine learning means feeding a lot of data to an algorithm, it learns from the data

No.
It means an algorithm which mutates based on the data inputs. What they're describing is very similar to what companies like Pixar have been doing with live-action motion capture applied to computer models. You rig up an actor with sensors and capture his motion, then apply those motions to the computer model instead of having somebody hand-animate the models.

What they're doing is slightly more along the lines of machine learning, but it's still a bit of a stretch to call it that. What they're doing is generating the mathematical models for character animation by using motion capture, instead of entering them by hand. The key difference between this and actual machine learning is that the algorithm for creating the mathematical model does not change... it's just the resulting model which is generated that changes. True machine learning would mean the algorithm which creates the models mutates and becomes better at creating the mathematical models, and that isn't what they're doing. In fact, they go so far as to mention that it requires actors to be highly skilled in their art- a true machine learning algorithm would, over time, get better at reading the actors thus requiring less and less skilled actors, until it reached the point where it could develop the same resulting math models from any average person with no acting skills at all. And they aren't even trying to do that.

Re:Question (1)

glebovitz (202712) | about 2 years ago | (#40984399)

No. It means an algorithm which mutates based on the data inputs.

  I thought it was a branch of computer scientific concerned with the design and development of algorithms that allow computers to evolve behaviors based on empirical data, This does not imply that the algorithm mutates, though that is one way of allowing the computer to evolve.

Re:Question (0)

Anonymous Coward | about 2 years ago | (#40988717)

Your definition does not match the commonly used definition, sorry: http://en.wikipedia.org/wiki/Machine_learning

There's a lot of practical stuff to machine learning - I don't think that can be said for self-mutating algorithms, not as the state of the art is.

Re:Question (4, Informative)

hazem (472289) | about 2 years ago | (#40980667)

A non-machine learning method would entail specifying a set of reactions for the set of inputs. If A then B, and if C then D... the hardcoding the relationships between the inputs and outputs. Machine learning, on the other hand, involves the computer program going through a pile of data and determing the relationships between inputs and outputs.

This problem is probably well suited to the technique because there are so many possible inputs and many nuanced outputs. In that situation, it's difficult if not impoossible to construct a programatic flow-chart that will perform well.

There's tons of material on the web now to help teach about these methods, like Coursera.com's courses in AI, Machine Learning, and even Probabalistic Graphical Models..

Re:Question (1)

zbobet2012 (1025836) | about 2 years ago | (#40981495)

Correct, wish I hadn't spend on all my mod points today.

Re:Question (0)

Anonymous Coward | about 2 years ago | (#40982195)

Well, it's still nonsense, nuances or not.
In games the characters should be controlled by the game unless you want to make some sort of parrot simulator.
The REAL question is not wether we can analyse human behaviour to death but wether we can make a good game using this technology.
For one, i'm not looking for deep human-level interaction in games per se.
The main character could be a pixel if the game is good.
These people would be the same people that would say that very good physics simulation makes great games.
They are not gamers.
They are scientists trying to get funded for playing with gadgets.

To be honest, i don't see the fun in ever more realistic games.
I don't want to play a real life simulator.
I play games to escape reality but more and more games feel like chores and refleting the boredom of everyday life.
I have no doubt that this technology will bring nuance to the virtual co-workers that are certain to pop up in the soon to be announced Office Workday Simulator.

Answer (2, Informative)

Anonymous Coward | about 2 years ago | (#40980675)

It presumably requires machine learning because the inputs (all possible permutations of bodily motion) are so diverse that it becomes pretty much impossible to hand engineer a decent response to them. Therefore you feed some training data (motion capture of actors) of paradigmatic or common inputs and let the algorithm learn how to respond to the myriad other inputs that you haven't provided. The situation is exactly analogous to handwriting recognition algorithms that the post office uses to sort and route your snail mail. There are a zillion ways to write the letter 'a', which makes a rigorous, engineered solution all but impossible, but you can feed a whole bunch of different examples of written 'a's to a learning algorithm and it will get really damn good at recognizing all the unseen, novel ones.

Re:Answer (0)

Anonymous Coward | about 2 years ago | (#40980941)

dont think so,if u build "machine learning" you are smart enought to figure out the 'complex" you described.

Re:Answer (1)

Ginger Unicorn (952287) | about 2 years ago | (#40982847)

Your gut feeling about the relative complexity of these two approaches is so off-base, that the only conclusion is you have no experience of this and don't have a clue what you're talking about.

Re:Question (1)

nancyfromafrica (2706953) | about 2 years ago | (#40981553)

We will get some awesome games with better AI in the future

Amazonin achievement (-1)

Anonymous Coward | about 2 years ago | (#40980453)

I hear the algorithm could detect every single nuance of body language, as confirmed by the asocial nerds developing the thing.

A better summary (1)

Anonymous Coward | about 2 years ago | (#40980857)

So, to cut past this article not knowing how to write, this is what's happening: A set of motions are captured for two people, one correlated to the other. So X is correlated to Y. Then, when gesture X is exhibited by the player, animation Y is done by the in game character. This is just coupled by the age old branching tree structure for AI to get some silly interacting dancing man.

It's not really interesting, it's just: When player moves X choose to play mocapped animation Y(z) where z is random (or whatever). I'm not really sure what they even intended here.

I want a game that understands sarcasm (1)

Anonymous Coward | about 2 years ago | (#40980889)

Now that that's a really useful invention....

Who is abusing moderation? (-1)

Anonymous Coward | about 2 years ago | (#40980935)

"Is this censorship?
We're not technically deleting anything. In fact "We" technically aren't really doing much at all. The masses are doing this for themselves (in theory anyway). And you are always given the option of clicking the threshold control over to '-1' and reading everything uncut, so I really have a hard time saying this truly is censorship. But if you really want to call it that, I can't really argue. We're trying to make as many people happy as possible here- if you don't like something, you can probably change it in the user preferences to more suit your tastes anyway."

Not true. "We" are throttling criticism of a few select editorial moderators.

"IP Restrictions: No single IP can post more than X comments per story. While this has problems, it would help prevent the occasional moron who likes to come in and post a dozen comments in a row on an article that really don't say anything. This could take a variety of shapes: No single IP being allowed to post more than 10% of all comments in a subject? A hard limit of 10 comments per IP per story? 5 minute delay between posts from any IP? Each method has ups and downs, but would probably solve the problem. The problem is that it would cause other problems so I'm not really planning on implementing this yet, but if I do, it will be fairly lenient."

Not true. This is very much implemented. It is used to prevent criticism of a few select editorial moderators, by silencing this criticism as well as replies to this criticism. E.g. replying to a -1 is by default difficult to due.

Re:Who is abusing moderation? (1)

Zaelath (2588189) | about 2 years ago | (#40980955)

I'm not sure what your thesis is, or who you are quoting, but whatever you're trying to do is sadly lacking in enough formatting to make it clear. Perhaps you could get some assistance from the APK jackasses in how to use the site.

Duck Hunt (2)

KalvinB (205500) | about 2 years ago | (#40981129)

Can we finally kill that stupid dog by flipping it off?

Pedantry (1)

grammar fascist (239789) | about 2 years ago | (#40981175)

...responding to the player's own body language rather than mathematical rules.

I've done a lot of machine learning. It's very much mathematical rules. The inferred rules just happen to be so complicated that it's often hard to recognize them as such.

Maybe the submitter meant "hand-crafted decision tree".

Re:Pedantry (1)

fractoid (1076465) | about 2 years ago | (#40981213)

Yeah, that bit bugged me too. It's like saying that a walking robot "responds to its environment rather than physics."

Re:Pedantry (1)

wermske (1781984) | about 2 years ago | (#40981337)

I understand the confusion. I believe context and choice of assumption is important in this situation.

The source article opens by indicating design/development is the area where advance has been made... "shows a new approach to designing video game characters that can respond to our body movements and body language." While they speak to "having to think about mathematical rules," there is no indication that such rule would be or are precluded from a mature model.

The innovation implies the challenge, "how to capture and incorporate subtlety and natural movement" -- we are expected to assume not that mathematical rules are bad...merely, in and of themselves, inadaquate to the challenge.

Re:Pedantry (0)

Anonymous Coward | about 2 years ago | (#40982337)

...responding to the player's own body language rather than mathematical rules.

I've done a lot of machine learning. It's very much mathematical rules. The inferred rules just happen to be so complicated that it's often hard to recognize them as such.

Maybe the submitter meant "hand-crafted decision tree".

Nope. ML means the algorithm for generating the decision tree mutates based on the inputs. This is a static algorithm which generates complex decision trees based on the carefully selected inputs of skilled artists. The algorithm doesn't learn, it has to be refined by the programmers, by hand. It's neat, don't take my post the wrong way, but it's a far cry from developing an algorithm which can mutate over time to generate better and better decision trees.

Wait a damn second!! (1)

moniker127 (1290002) | about 2 years ago | (#40981417)

Understanding of human interaction? Programmers?

Eh? (0)

Anonymous Coward | about 2 years ago | (#40982055)

"Machine learning" without "mathematical rules"? How exactly do they manage that?

So next time I get angry: (0)

Anonymous Coward | about 2 years ago | (#40982771)

So next time I get really pissed off, the console will know beforehand that I'm going to throw my controller at it?

Great (0)

Anonymous Coward | about 2 years ago | (#40983079)

This is great for me with aspergers, now videogame characters can missunderstand me as well.

magic instead of math? (1)

shadowrat (1069614) | about 2 years ago | (#40983587)

responding to the player's own body language rather than mathematical rules

I confess to not reading TFA, however, I have yet to see machine learning that doesn't rely on mathematical rules.

Check for New Comments
Slashdot Account

Need an Account?

Forgot your password?

Don't worry, we never post anything without your permission.

Submission Text Formatting Tips

We support a small subset of HTML, namely these tags:

  • b
  • i
  • p
  • br
  • a
  • ol
  • ul
  • li
  • dl
  • dt
  • dd
  • em
  • strong
  • tt
  • blockquote
  • div
  • quote
  • ecode

"ecode" can be used for code snippets, for example:

<ecode>    while(1) { do_something(); } </ecode>
Create a Slashdot Account

Loading...