Just as a continuation of my previous concerns about maintaining content throughout the game. In a short demo you could potentially have 8 or 16 pre-determined conversation paths where at each point in the conversation the system was able to identify 8 or 16 statements which each have a scripted pre-recorded response; and with significant research you could ensure that 95% of people’s interactions with this short demo would have a seemingly natural flow (regardless of whether it is scripted or not). The problem is that you have just produced a couple hours worth of conversation and scripted interaction for a 5 minute demo.

While it is entirely possible that you could take this approach with an entire game if you were willing to invest hundreds of millions of dollars to create the most realistic virtual-boy simulation, in reality I think it is unlikely that they would be able to afford to produce a series of conversations which were (much) more complex than the typical RPG; and it is entirely possible that they might have to give you similar prompts about what to say in order to keep you from having the simulation respond with garbage most of the time.

I do admit that while I have worked with voice based systems in the past, my knowledge of cutting edge techniques is limited so there may be research and/or tools which would reduce the work I would anticipate this would take; but I still think that managing the interactions of a virtual boy might be as complicated as producing many of the biggest budget games today.

Tell me what speech synthesis software even comes near to the quality of Milo's speech...

I've seen many and they're nowhere near as good. Yes, they are usable for blind people to read web pages, but they don't sound like a real person reading, much less a person talking with emotion (which is what Milo sounds like).

For me most of what i seen so far about natal in action (real action not trailer) has been lame cus it has a lot of leg so far , Milo was impressive for me in the moment i saw it on the press conference just like was natal , with those Sci-Fi trailer its easy to make it looks like the future , but when you see it  in action you see the legs , and wonder , "How a leggy thing like this make all that cool stuff i saw on the video ?" well , i tell you what , so far for me Natal is a over hyped EYETOY but i open to see what MS has to show us at E3 and if it does everything that is does in that trailer i will buy one for sure (i mean for sure if nintendo does not come up with something cooler to take my money away from MS...I think 3ds will make that happen) 

This points to the biggest issue with Natal.

There is so much misunderstanding of what it can and will do. And overselling (which Microsoft has a reputation of doing) could ultimately be a negative on this hardware. In other words, not everyone realizes that the AI is a tech concept video and not actual "gameplay footage."

In other words, there are a lot of people who expect it to do much more than it will be able to do. Not that it will not be a good technology, but it may not be the technology that some people (incorrectly) expect.

Mike from Morgantown

Agreed. To this date we don't even know if it can do something like Wii's IR aiming.

With such basic unknowns, no one can really predict how successful it will be.

It has an accuracy of 3mm in the XY dimensions at 2 meters from the camera.  Plenty of accuracy for using your hand and an on-screen indication.  For what purposes were you wanting Natal to have something similar to IR aiming capabilities?

For the same purposes that the Wii uses it (or more). Isn't it kinda obvious?

Unfortunately, I do not currently own a Wii. 

Well, knowing that Natal provides 3mm of accuracy, does that suit the purposes?

~BillyBOB

Do you mean that it can tell if I move a finger left by 3 mm? In that case, pixel-accurate aiming at 1280x720 would require more than 3 meter movements, even if you make that 10-pixel accuracy that still makes it more than 30 cm, looks like way too much.

Just speculation which is probably wrong...

I don't know anything about speech synthesis, so I didn't make any comments on it. I was more interested in talking about speech recognition and gesture/emotion recognition, both of which I believe are possible for Milo. Though, my argument works against myself, and while I might know a bit about speech recognition, I am by no means an expert.

But since we're on speech synthesis, I don't think that's what they're trying to do with Milo (yet). I do think that what Milo is saying is scripted, but I don't think the experience is scripted. That can go towards the illusion thing Molyneux talks about, and that's okay so long as the experience itself is immersive as a result.

This is obvious but we won't really know until it's released or shortly beforehand.