this is pretty obviously crap to anyone with any real background in AI. even giving them the benefit of the doubt and assuming it's not scripted, it's so painfully obvious that they're resorting to cheap mind tricks instead of attempting anything remotely close to real AI.
first of all, you'll notice that milo's first couple sentences are just repeating back what the user said. Molyneux claims milo can recognize emotion in peoples' voices. Sure he can, just as long as they're saying things like "I FEEL A BIT NERVOUS." wow, she said she was nervous and he figured out she was nervous somehow! voice recognition software isn't that innovative. never mind that she clearly ISN'T really nervous, because MILO certainly doesn't.
Milo's next reply is "Thousands of people..." kudos on making it sound like a thoughtful response, but it's just repeating what the user just said. <a href=http://en.wikipedia.org/wiki/ELIZA>Joseph Weizenbaum figured that trick out in 1964. not much innovative there either.
the rest of Milo's tricks work pretty much the same way. when the program picks up on "homework", instead of coming up with a vacuous statement, it instead has milo look quiet and sheepish. it's a fairly safe bet that regardless of context, any kid would react the same way when someone brought up homework. also notice how milo completely ignores the part about being beaten in football before. is there any kind of memory at work here? any learning algorithm?
apparently milo's "homework" is drawing fish in a journal. that seems fairly vague, but ok. apparently he also has no idea what fish look like despite standing next to a pond full of them for god knows how long. as soon as it recognizes neat little eye toy goggles trick, the game automatically positions the camera over the pond and initiates the fish-grabbing minigame, which would be a lot more impressive if johnny chung lee hadn't proven the concept with a wiimote two years ago. the water effects on the reflection do look pretty good though. actually the animation as a whole is pretty good, so good job on that i guess.
"they're only fish, but they're trickier than you think." isn't this an oddly vague thing to say about drawing? it sounds naturally enough if you were trying to catch or cook one. what's so tricky about drawing a fish? especially if, like milo, you're supposedly good at drawing? why wouldn't you say "i have trouble drawing fish, for some reason." why won't milo refer to the object and the action in the same sentence? because it's easier to pick out one word, and have a generic statement in response to that.
claire then graciously volunteers to do milo's homework for him and draws a fish. what if she had drawn a bird? or a house? or (let's be realistic here) a cock? chances are milo would react no differently. it's hard to hear over molyneux's self-aggrandizing voiceover, but it sounds like milo limits his response to the color of the drawing. it's nice that you got a program to recognize a piece of paper and any colors on it, but art is so subjective that it's practically impossible to get a computer to recognize anything beyond simple geometric shapes. (even getting a computer to recognize printed text, regardless of font, would be a huge breakthrough.)
why do you think "only a select few" will actually get to try milo at E3? it's because any in-depth attempt to really communicate with milo will expose the thousands and thousands of flaws in the supposed "AI". Basically natal is just a ripoff of the wii and eyetoy with some voice recognition thrown in, and molyneux slapped it on a half-assed AI. This is what apparently passes for ground-breaking innovation at microsoft.
first of all, you'll notice that milo's first couple sentences are just repeating back what the user said. Molyneux claims milo can recognize emotion in peoples' voices. Sure he can, just as long as they're saying things like "I FEEL A BIT NERVOUS." wow, she said she was nervous and he figured out she was nervous somehow! voice recognition software isn't that innovative. never mind that she clearly ISN'T really nervous, because MILO certainly doesn't.
Milo's next reply is "Thousands of people..." kudos on making it sound like a thoughtful response, but it's just repeating what the user just said. <a href=http://en.wikipedia.org/wiki/ELIZA>Joseph Weizenbaum figured that trick out in 1964. not much innovative there either.
the rest of Milo's tricks work pretty much the same way. when the program picks up on "homework", instead of coming up with a vacuous statement, it instead has milo look quiet and sheepish. it's a fairly safe bet that regardless of context, any kid would react the same way when someone brought up homework. also notice how milo completely ignores the part about being beaten in football before. is there any kind of memory at work here? any learning algorithm?
apparently milo's "homework" is drawing fish in a journal. that seems fairly vague, but ok. apparently he also has no idea what fish look like despite standing next to a pond full of them for god knows how long. as soon as it recognizes neat little eye toy goggles trick, the game automatically positions the camera over the pond and initiates the fish-grabbing minigame, which would be a lot more impressive if johnny chung lee hadn't proven the concept with a wiimote two years ago. the water effects on the reflection do look pretty good though. actually the animation as a whole is pretty good, so good job on that i guess.
"they're only fish, but they're trickier than you think." isn't this an oddly vague thing to say about drawing? it sounds naturally enough if you were trying to catch or cook one. what's so tricky about drawing a fish? especially if, like milo, you're supposedly good at drawing? why wouldn't you say "i have trouble drawing fish, for some reason." why won't milo refer to the object and the action in the same sentence? because it's easier to pick out one word, and have a generic statement in response to that.
claire then graciously volunteers to do milo's homework for him and draws a fish. what if she had drawn a bird? or a house? or (let's be realistic here) a cock? chances are milo would react no differently. it's hard to hear over molyneux's self-aggrandizing voiceover, but it sounds like milo limits his response to the color of the drawing. it's nice that you got a program to recognize a piece of paper and any colors on it, but art is so subjective that it's practically impossible to get a computer to recognize anything beyond simple geometric shapes. (even getting a computer to recognize printed text, regardless of font, would be a huge breakthrough.)
why do you think "only a select few" will actually get to try milo at E3? it's because any in-depth attempt to really communicate with milo will expose the thousands and thousands of flaws in the supposed "AI". Basically natal is just a ripoff of the wii and eyetoy with some voice recognition thrown in, and molyneux slapped it on a half-assed AI. This is what apparently passes for ground-breaking innovation at microsoft.
