By using this site, you agree to our Privacy Policy and our Terms of Use. Close

Forums - Microsoft - Exposed: Natel Milo smoke and mirrors by Peter Molyneux

Reasonable on 07 June 2009
Procrastinato said:
Milo is using a knowledge-based system to recognize words, and symbols. That's not a "hoax", its textbook AI. Nor is it revolutionary, I might add. Far from.

Honestly people are only excited about this because the vast majority of the media has ignored the accomplishments of AI research over the years, or believed it to be pie-in-the-sky and the stuff of science fiction plots. This isn't science fiction, nor is it really very interesting. KBS's were the very first form of AI. We're talking the 1950s here.

Very good point.  Heck, I remember devlivering my Thesis on Expert Systems in the eighties, where the use of such techniques was part of the interface elements for medical expert systems and the like - i.e. it interprets responses, checks against database key words, etc. to produce a diagnosis.

You have to applaud the marketing folks behind the Natal presentation, as they sure got the column inches they were no doubt aiming for.

 



Try to be reasonable... its easier than you think...

Around the Network
heruamon on 07 June 2009
drakesfortune said:
I have to agree. Natal looked terrible to me. I don't understand how so many people seem fooled by what this thing can do. They made a LOT of bold claims, but the reality is they didn't prove anything with Milo, which was displayed in a controlled way, and the ball demo shows SO MUCH lag that it'd be no fun to play. In fact, lag is going to be the thing that makes Natal useless.

I think there's absolutely no way that microsoft can deliver what they're promising with Natal. It's a souped up eyetoy, and when you really look at what they showed, with the horribly inaccurate paint demo, and the lagalicious dodge ball demo, there's no reason to think this is anything more exciting than eye toy 2.0. This looks less accurate than the wiimote 1.0 and with much more lag. Do not want.


10 post...named Drakesfortune...lol...anyway...yep...it's just Eyepet...and when it cause Xbox 360 to open the gap to 10 million+ between PS3, it will be another marketing smoking mirrors of M$...



"...You can't kill ideas with a sword, and you can't sink belief structures with a broadside. You defeat them by making them change..."

- From By Schism Rent Asunder

JaggedSac on 07 June 2009
MikeB said:
JaggedSac said:
MikeB said:
From MTV multiplayer:

"So what was technology and what was a trick? It appeared that Milo’s ability to recognize actual words is completely AWOL at the moment. Former MTVer Stephen Totilo joined me for my meeting, and Stephen’s attempts to interact with Milo resulted in mere quaint smiles and frowns. Molyneux went on to say that players must “train Milo to their voice.” That may be, but based on this demo there’s still a lot of ground to cover."

So it appears the technology (at least at this point) is not nearly advanced enough to allow a random user to interact (for example a gaming friend). Possibly bad articulation or dialect will have an affect on the effectiveness of this technology, let alone if people want to play in their native languages like Japanese, German, French, Italian, Dutch, etc this may pose a far greater hurdle to overcome.

I think they should have waited a couple of years until they really have something to show the public.

So do you not like the idea of training Milo to understand things?  EndWar had vocal commands, was that not implemented for non english countries?  If it works for that game, which has already been released, it seems quite feasible for it to be used in this game.  This would just be a rather large extension of those methods.  In this case, a huge dictionary of terms linked together rationaly to provide the illusion of sentience.  Vocal recognition is nothing new and that seems to be the hiccup you see here.  I am not seeing it however.


No, I would not like to train a Milo. I want such a game to be fuss free and I want to be surprised. If I have to train him the commands, I rather train my cat instead.

I remember an old Dutch languaged text adventure form the early 80s which blew me away with its level of word recognition. I was a preteen boy and there was an old lady, for fun I said fuck that old lady and the game responded, he was not going to do that to such a poor old lady. That's the kind of suprise response I am looking for.

So the smoke and mirrors part was the training aspect of the game?

And of course an old text based adventure will not need training.  The token was clearly input and no interpretation was necessary.  That is a lot different than vocal recognition.  The training is to help the program understand how you speak, drawls, lisps, etc.  Windows does the same thing, it has you speak paragraphs into the mic, the more you speak the more accurate the training.  This is also what EndWar did.  It is a necessary evil with vocal recognition due to the differing vocal patterns of humans, at least at the current time.  But I do not see it as a deal breaker.

The problem is in no way in how to use the token once it is successfully deduced, it is in how to gather what the tokens are.



D_Boy on 07 June 2009

If it is indeed sentient AI, it will realise that the best course of action would be to send a T-800 back in time to stop Microsoft from ever showing Project Natal at E3.



heruamon on 07 June 2009

I'm just think back to last E3...and reflecting on that towards 2010 E3...it's goign to be so interesting to see how it all gets rationalized...



"...You can't kill ideas with a sword, and you can't sink belief structures with a broadside. You defeat them by making them change..."

- From By Schism Rent Asunder

Around the Network
ramses01 on 07 June 2009
drakesfortune said:
I have to agree. Natal looked terrible to me. I don't understand how so many people seem fooled by what this thing can do. They made a LOT of bold claims, but the reality is they didn't prove anything with Milo, which was displayed in a controlled way, and the ball demo shows SO MUCH lag that it'd be no fun to play. In fact, lag is going to be the thing that makes Natal useless.

I think there's absolutely no way that microsoft can deliver what they're promising with Natal. It's a souped up eyetoy, and when you really look at what they showed, with the horribly inaccurate paint demo, and the lagalicious dodge ball demo, there's no reason to think this is anything more exciting than eye toy 2.0. This looks less accurate than the wiimote 1.0 and with much more lag. Do not want.


The lack of vision of so many people on the net is always so disappointing, yet of course completely expected.  If your mind doesn't just explode with the possibilities of NATAL, then I feel sorry for your lack of imagination and creativity.  MS has done what they do best, create an interoperating suite of functions where the value of the whole exceeds the sum of the parts.  1 aspect of NATAL is a souped up Eyetoy, the visual spectrum camera.  To that, MS added a depth perception camera, a sophisticated microphone system, and most importantly, custom processors and software to integrate the data and allow facial and voice recognition as well as full body motion capture.  People are overlooking the importance of the backend systems here. 



Seece on 07 June 2009
Akvod said:
Squilliam said:
Akvod said:
daroamer said:
JaggedSac said:
LOL. What was deceptive about it? The voice recognition, facial recognition, etc? Did people actually think they had developed the most sophisticated AI of all time?

Exactly.  Plus, Peter himself admitted the demo was using a lot of "tricks" to be able to do what it did.  That doesn't mean it was deceptive.

But if you don't admit those tricks on the spot (E3 conference), then it is deceptive.

Magicians don't need to admit they used tricks because it's already assumed that they are. In this case, it's almost like a person doing something that seems magical, and not admitting that it was only a magic trick in the end.

Yeah,  the whole 3d graphics rendering you see in games is one big deception. I expect you to now go off on one big thread about how game artists and programmers have been deceiving you.

Your thread awaits, go to it!

 

I'm not reffering to the 3D graphics rendering (if you read my previous posts in the same thread), but how the presenters implied that Milo has emotions, and is an actual sentinent AI.

Please don't put words in my mouth, and use straw mans.

Thats bloody rich coming from you. Why are you making out Milo is something seedy? You need to have that sort of mindset first to think that way ......



 

Words Of Wisdom on 07 June 2009
HappySqurriel said:

Come on, Peter Molyneux would never promise something that he couldn't deliver! Just look at Black & White, we all know after he demonstrated highly trainable sentient ai with that game it rapdily spread to every game and we've never dealt with scripted AI again. On top of that Project Ego (Fable) really was the greatest RPG of all time, and every decision you made as a young boy had massive consequences as you aged.

Takes one to know one. 



bardicverse on 07 June 2009

I wonder how many people from Nambla are going to get Natal for Milo and try to get him to strip for them LOL



Procrastinato on 07 June 2009
Reasonable said:
Procrastinato said:
Milo is using a knowledge-based system to recognize words, and symbols. That's not a "hoax", its textbook AI. Nor is it revolutionary, I might add. Far from.

Honestly people are only excited about this because the vast majority of the media has ignored the accomplishments of AI research over the years, or believed it to be pie-in-the-sky and the stuff of science fiction plots. This isn't science fiction, nor is it really very interesting. KBS's were the very first form of AI. We're talking the 1950s here.

Very good point.  Heck, I remember devlivering my Thesis on Expert Systems in the eighties, where the use of such techniques was part of the interface elements for medical expert systems and the like - i.e. it interprets responses, checks against database key words, etc. to produce a diagnosis.

You have to applaud the marketing folks behind the Natal presentation, as they sure got the column inches they were no doubt aiming for.

 

Indeed. They dressed up an exercise in classic AI systems, using modern-day storage mediums and modern input devices, and managed to pass it as the "next big thing" in gaming.

As a computer scientist, I applaud this kind of stuff, since AI research has always been in the realm of the not-very-practical...  but for the same reason, I also know how flawed its likely to be, as I'm certain anyone with a degree in CS/CSE/whatever-your-school-called-it, and any computer vision, pattern recognition, or AI coursework under their belt (let alone research) does as well.