E3: Molyneux and Milo Comments by Ellie Gibson

Got to love this guy's enthusiasm (and Ellie, of course).

That said, like all "artificial intelligence" it will most likely seem intelligent for an hour, and artificial after that.

I actually laughed out loud at the Times New Roman joke :(

"The fact that Ellie told him a joke and he laughed is proof enough that it works to me."

Hmmm. All that really proves is exactly what they described, that it can recognise tone of voice. We can't then assume that EVERYTHING works just as we imagine. And it is about what we imagine, because actual statements about what this software does as a whole are absent as far as I can see.


@Negotiator

Ok, something weird is going on here. You are just too damn enthusiastic and believing to be a normal person. You seem utterly convinced in every way that this thing is one step short of Skynet, yet you have seen exactly the same highly staged demo (followed by some extremely brief and simple voice tone recognition) as the rest of us.

PM said himself the voice tech is lifted from Windows. And Windows voice recognition is not the best in the business (I don't just mean the games business btw). So why are we suddenly assuming Lionhead have the best AI in the world, just because they have skinned it in posh graphics?

Do you work for Lionhead or something? Or are you genuinely just a massively blinded fan? This demo has proved almost nothing to anybody even remotely interested in somethign called "evidence", so it is ludicrous to start forming firm conclusions either way.

knob or nob

I'm impressed, I genuinely am, but this isn't for me:

1) Waving my arms around like Magnus Pike to control things wore off quickly on the EyeToy.

2) I already groan at the DS games that require me to speak into the microphone. Frustrated expletives aside whilst in the fray, I really have no interest in verbally interacting with my telly.

3) How will the facial recognition cope with all the ugly gamers in the world? :)

@Retroid [mod]

I think perhaps the issue some poeple have is that this feels like the same cycle they have seen before. PM being open about the tricks being used may speak volumes for his honesty, but it doesn't make anyone any more sure that the end result (you know, the actual game) will include all of the features being discussed at this stage.

Now I thought Fable 2 was a great game and I enjoyed playing it a lot. But I also recall PM talking about LOVE being one of the three USPs (the other two being one button combat and something else, the dog perhaps). And that turned out to be utter bollox. Now maybe Milo will end up being an entertaining game, but I and others are struggling to see the connection between that end result and the tech demo we have seen so far.

Though deep down I'm afraid that Milo will in fact work every bit as well as they say, will become wildly popular and will indirectly lead to the death of deadpan. Jokes are always funnier with no inflection at all - when they're told as stories, not as jokes. But this is a truth that poor Milo will never be able to understand. This is actually rather upsetting.

So this is just lovely, I buy this game and I'm going to have an adorable little kid telling me I seem sad every time I look at him. And I will never be able to tell him why.

Does this software understand linguistically what we are saying or is it responding to wave pattern analysis of our voice?( the feeble joke example ). It seems cleverly put together but at the end of the day, its not intelligent enough if it only has 500 words, and especially if it doesn't understand the words in context. I bet you could fool the shit out of it by saying things but make sentences sound like questions.

Smoke and mirrors is all i see with this tbh.

@Darren

LCP was ace!

feed him dog food and make him dance to records all day...ahh the good times :P

All the kid's responses sound like pre-recorded soundbytes rather than actual speech snythesis - if this is the case he'll have a very limited set of audio responses making any real notion of AI redundant. Also means he won't be able to greet me as "Fidjit" in the same way he greeted the girl as Claire.

Also, voice recognition and appropriate response in games is nothing new. Worked a treat in Rainbow Six 3 on the original Xbox. ( although not with intonation/inflection admittedly)

@Negotiator

Ironic name given what you write. Anyway..

If you're trying to say that this hasn't been driven forward hugely up Microsoft's agenda because of the Wii then you live in a different world to me. Watch the promo reel and tell me that the pictures of the whole family sitting on the sofa and having so much fun in front of a games console (did they even show a 360 itself?) hasn't been heavily influenced by Nintendo's own adverts which have done them so well.

CyberCop 'If your an adult playing Milo then your playing it for the wrong reasons'

"EA have hacked Natal wheel-free Mario Kart Wii style controls into Burnout (presumably paradise), and motion sensed foot control for pedals, too."

There's much more to it than that though. Reading simple movements and translating that into control input is the easiest part, as it has been done already by the likes of EyeToy. But working facial recognition and the more elaborate stuff into a game is much more tasking.

I don't think its as straight forward as saying "Reading simple movements and translating that into control input is the easiest part".

It might be easy to do the first bit, i.e reading the simple movements.

But translating those simple movements into usable effective controls is far from trivial (as we have seen so with so many damn awful 3rd party Wii titles, and a handful of similarly poor SIXAXIS implementations).

EyeToy games always asked very little of the accuracy of the player input, and that was their strength. The same can't be said of some of the examples in the promo movie (like the kungfu game for one).

It's the beginning of Halo, I tells you!

Great interview Eillie, all the right questions, that normally people wouldnt ask :D

For people who are saying wow it laughed thats amazing, thats showing AI.. its not really. The voice recognition software sees if it can understand a percentage of the words, if not, make Graphic laugh/giggle depending on the percentage.

Hes already admitted that the voice activation is based on Windows, and that my friends shows how well scripted this was... Voice technology in Vista is not that good.. trust me Ive used it.. its quicker to phone someone else up, get them to type it and then email it across.

I will certainly eat my hat if this comes off, but i reckon its another 15-30 years before we get what people are expecting, and what they will actually get.

Wow it's amazing to witness the narrowmindedness of some people.

Have you people actually watched the press conference? Fair enough, they went on about Milo a bit even though that WAS fing cool. The point is, it doesn't matter WHAT they showed as a demo for the new camera, it matters that we were shown how accurate it can be.

Anyone that has watched the videos knows this, and comparing it to eyetoy is unbelievably dumb. If I was Nintendo I would be very worried right now because this actually makes Wii controls look pretty damn stupid tbh.

"It might be easy to do the first bit, i.e reading the simple movements.

But translating those simple movements into usable effective controls is far from trivial"

I was referring to the driving controls example. That's a limited set of simple movements resulting in relatively simple control inputs for the driving game. There's a reason that that's one of the first 'real use' examples.

what a sarcy cunt ellie, thats not how you are meant to interview people.

@Collymilad did you even watch the video trailers for the product.. the kid playing Godzilla, the monster moves his feet before the kid does... smoke amd mirrors :D

Cool idea, not going to work as they are presenting it in reality.. which is a shame, but also a big lie to steal the show.

"Cool idea, not going to work as they are presenting it in reality.. which is a shame, but also a big lie to steal the show."

True. Will probably go the way of the amazing table PC...

What I think is very interesting is this is quite similar to the type of virtual person various researchers in AI have been putting together for the last few years. If you look at those attempts none of them are truely AI or even remotley close but they are all aiming for roughly the same thing as Milo with giving realistic responses to various types of human input. I find it fasinating that we are not too far away from being able to have something like that on domestic technology now.

As to if its a game or not well I don't see it being something I would pay money for as I would think the novelty will wear off pretty quickly once you guage just how limited it really is but as a glimpse at what future GUI's might start to become its pretty dam cool.

The implications for this tech in role playing games has really got me thinking.

In a few years time, once the tech has improved and been put into games, it will have enormous impact.

Say it's 2014, I'm playing Elder Scrolls 7, I reach a tower to save the beautiful princess, the screen cuts to her, she runs forward looking at me, actually looking at me due to the facial recognition and she begins.."oh hero, you've come to save..."

And then she stops, looks at me more closely and screws up her face in disgust.

Because if I'm playing this late at night, chances are I'm sat in my favourite easy chair wearing just my pants and eating snack food.

It boggles the mind.

"The implications for this tech in role playing games has really got me thinking."

What about poker games? In a couple of years you'd need a good poker face or else start wearing sunglasses and silly hats to get away with bluffing... ;)

After all the shit I've been pouring on Molyneux (I doubt he ever reads it, but still), I do feel compelled to say that I actually quite like him. Not as much as Ellie, obviously, but still. Having grown up with games and now being a games scientist as profession, I really like how Molyneux is constantly looking into new ways to foster emotional involvement or other affective responses during gameplay. He's also being very honest about it in this interview, of course he didn't beat the Turing test, he's just made a demo to indicate which direction the new technology could take us.

The problem is that this is not usable for games in the near future and in that respect I don't think Microsoft should have included it in the press conference, but I can't really blame Molyneux for that. Likely, he will try to incorporate this into a game and it will be horribly botched. Actually I think Molyneux should stop making games and MS should just pay him as a video game visionary, creating demoes of exciting new directions, that other devs can then capitalize on. He'd have my application letter and resume within nanoseconds too.

@ rhubarbandcustard Yeah its actually quite funny because if they used something like this in an RPG you would actually have to act your part. I am not sure if I find the picture of me acting my part alone in front of the TV funny or not.

Presumably they will be able to get around the sitting in your pants problem though by just having it as a given that you are wearing a suit of armor. An unlockable pants only mode could be fun though.

i love ellie :(

A boy who only understands 500 words isn't that impressive.

But what about an animal that can understand 500 words? Now that would seem most impressive - to the masses.

MS speech recognition is extremely good once you have trained it properly.

Jeez, what is Ellie's fascination with drawing 'nobs'?

MS speech recognition is extremely good once you have trained it properly.

Not going to be much fun of a game if you have to stand there teaching it all 500+ words :D... yes i know you would do that when you set up the camera etc and not in game, but still.

"That's the interesting thing, you see. We've been very, very clever about this." Goes on to describe not-at-all-interesting-and-fairly-straightforward way of censoring players input

It would be quite easy to write a Molyneux hype AI I reckon :P

Obviously this is way too ambitious - even for Molyneux - but that's the beauty of it. Yes; it probably will 'fail' in more than one way, but I'm glad to see such a high profile videogame developer take tiny incremental steps towards radically improving computer A.I.
...and am I the only one who was impressed with Molyneux's relatively restrained manner (i.e. no wildly swinging arms about and describing this as the Second Coming)?

I don't think you guys are getting it.

With facial recognition and voice recognition and even just "faking it" this is a tech changing thing. Sometimes we just want someone to talk to - and if this is a learning AI then it will "learn you" as time goes by.
And this would be perfect for the elderly that often have no one to talk to - which is what they need. Someone to talk to. If a robotic dog can do wonders for them then this is life changing. So yes, this changes alot.

With voice, facial and hand writing recognition, I can see many teaching tools being developed with Milo / Nadal in mind.

English / French / Spanish ... language learning courses, with the "teacher" being able to recognize accents and voice inflections

Math / Physics / Science courses

Music courses with the teacher recognizing the placing of your fingers on a given instrument

Exercise routines being perfectly monitored by the game.
.........

Besides all the game aplications of using gesture / voice recognition in combination with the traditional controllers

Hi Milo how are you today?

Sorry mum told me not to speak to strangers!

You could maybe teach him that!