Tonight, I got an incredibly brief chance to test out Microsoft's Project Natal camera technology with Peter Molyneux's "Milo," along with three other E3 judges. In the brief time allotted to the group, we experienced many of the same moments shown during Microsoft's keynote earlier in the day: introducing ourselves to Milo, being recognized by Milo, taking a brief walk to the pier, donning a pair of goggles, and running our fingers through the water. Before any of us would interact with Milo, Lionhead's Claire – seen above in a still from the E3 Milo video – had a brief, but natural, exchange with Milo. It began with Milo recognizing her face and addressing her by name.
Once Claire's portion was over, we took Milo out for a spin. For my part, I walked with Milo over to the pier, grabbed the goggles that he threw at me, performatively snatching at the air, having already seen the video during the earlier keynote presentation. To put them "on" I had to make loops with my thumbs and pointer fingers and wrap them around my eyes. A guide on the bottom of the screen instructed me to make them just so. Goggles on, I leaned over the pier and splashed a reflection of myself in the water. I could drag a finger around or make a larger splash with my hand. I could even ripple the water by "lowering" my head towards the screen. And that was about it for my section (see above: brief!). But the most impressive thing I saw happened next.
Up next was Victor Lucas, host of The Electric Playground, who was asked to form a bond with Milo by walking up to him while he was on the swing and simply stating his name. "Victor," he said. He then walked back (to ostensibly reset the scenario) and, upon walking back up and being facially scanned, was greeted by Milo. "Hello Victor," Milo said, in his own voice. No stutter. No audible edit in the welcome. That's impressive!
Later, when trying to "converse" with Milo, Victor consistently stumped him by asking questions. To be fair, Molyneux warned us that this is where the demo would fall behind (and specifically said asking questions wouldn't get Milo to talk). With the pressure of being instructed to tell Milo a joke, Lucas (and, indeed, the rest of us) were unable to offer even a single joke. Molyneux was trying to show how Milo reads one's voice, and the act of telling a joke is distinctly readable, he explained.
Not interested in being friends with an 8-year-old boy? Molyneux said there will also be a "Millie" – a female equivalent. Whichever gender you choose, Milo (or Millie!) will form a relationship with you and, if you so choose, other members of your family. Molyneux asked us to imagine leaving Milo on the screen, and allow various family members interact with him. Milo would have a different relationship with each person, and would even reference you (or them) in conversation with the other. In other words, think of Milo as a family friend ... who just happens to live in a box on your wall.
Despite Molyneux's assertion that it incorporates over five years of development work on the emotional AI, "Milo" is still a very early piece of technology. However, at any stage of development, Milo represents an ambitious project, presenting a virtual relationship that suddenly seems very real. This isn't any old Tamagatchi. And that it's still so early only encourages me. As for when we may see whatever the "Milo" demo will become – and yes, Molyneux confirmed it will be a game "with a score" and all the other trappings of a video game – he simply said that he'd hope it would be available as soon as the Project Natal camera was on store shelves. Now, when that is, nobody's saying.
... We'll ask Milo.