The Origins of Molyneux’s Milo

molyneuxMost of the world met Milo for the first time last Monday at E3. I first encountered him about five years ago at GDC. That’s when Peter Molyneux introduced me to Dimitri, a young boy who was furiously working on his math homework in Peter’s hotel room. “Hello, it’s nice to meet you,” the young lad said before turning his eyes back to a ruled notebook.

Little did I know that half a decade later I’d meet Dimitri again — albeit this time in virtual form.

Molyneux is a big idea guy. And Milo (renamed from Dimitri for legal reasons) is his most ambitious and riskiest concept to date. It’s a natural outgrowth of the artificial intelligence-driven creatures in Black & White and the dog in Fable II. But those AIs were supporting characters in a larger game world. Milo (or Milly if you select a female) is the game.

What inspired Molyneux to build Milo? Dimitri, who’s now in college, sees Peter as a role-model. Over the years, as Peter spent more and more time with Dimitri, he began to reflect on his own childhood and the choices he made growing up. How would Peter’s life have changed if he made different choices? It’s a question we all ponder at some point.

Milo, I expect, will give us a chance to re-examine our own choices by guiding a young avatar through life. He will ask questions, seek advice, and, if the technology works, form a bond with the user.  And when I refer to the technology, I’m talking about the AI – not the Natal camera.  Milo has been in development for years and the camera support was only just added in the past 5 months.

Here’s what makes me the most curious about Milo: What limits will Peter place on the game’s choices and the character’s actions? It’s one thing to slap around a creature in Black & White to condition him, but what will the tolerance be for such extremes in a game like Milo? Peter has already said there will be limits.  But could your Milo end up being a criminal?  Could he mistakenly drown in that beautiful fish pond?  Or, will Milo go off and meet a girl (or boy) at school, live happily ever after, and forget to come back and visit you?

11 Responses to “The Origins of Molyneux’s Milo”

  1. Lazy Flamingo Says:

    Good read Geoff.

    The problem with A.I is that it’s always pre-determined, no matter how much time, effort or dedication you’d give into programming it, it’ll never be an entity of its own that thinks and reacts in different fashion. There’s always a pattern, some sort of predictability, they can’t break that, at least in today’s world.

    You’d ask Milo a question, it might answer you, perhaps you’ll raise an eyebrow for it, that’s cool, but can you ask him a chain of questions? each one linked to the other, then chain them with others, seems imposable. With that aside, all what’s left is the virtual-physicality interaction, and that all depends on how well Natal’s implemented, that remains to be seen. Hope it’s not smoke & mirrors though, really hope it’s not.

  2. DavisDr Says:

    Man, this sounds wierd as hell. Wait, is Dimitri some kid and what does he has to do with everything? You should do a whole five-part Bonus Round with Molyneux with him explaining this stuff. Is Milo an RPG? Some mad pedophile dating sim? Is it a tech demo…wait, who is that Dimitri guy?

  3. HakarSidik Says:

    first of all i would like to apologize for my english since im not a native english speaker.

    In what ways is MIlo a game though? I mean, im very interested in the concept, but to who are the actually planning to sell it too( if it is being sold) and if they are not selling it, why show it in a microsoft game presentation?

    Even though im very intrueged i have to say that these kind of developments are also scary because, as you stated, what are the limits? (plot of terminator movies anyone?) for all i know AI could be taking over computers if more Milo’s show their faces..

    Im very interested in hearing more but dont tell me i didnt warn you about robots taking over and Arnie coming to save us.

  4. Rob "Azuma" Kimball Says:

    Milo is impressive…and creepy. So did you get hands on time with it? Was it as interactive as Molyneux makes it out to be? Not that I would have any reason to doubt him lol…

    I wonder about how scripted the events are…I picked up on a few things in the demonstration with Claire. Namely the part where Milo throws the goggles. I noticed Molyneux said that everyone reaches for them like they are really there. That leads me to believe that many people have experienced this event. Is Milo’s progress more on rails then it appears in the video?

    Between my time as a salesman and my debate class in school (for game design), I learned a lot about the art of conversation. You can lead people in directions you want without them knowing it. I know the AI has to do this to some degree, but how much I wonder?

    I know this is a huge leap in technology, but I can’t help but think of Seaman from the Dreamcast days. At least Milo isn’t as creepy as the fish/man/frog thing.

    I doubt Molyneux would be stupid/insane enough to allow you to beat the hell out of Milo like in Black and White. Anyone with common sense or a shred of moral decency would understand the consequences of that choice. The ESRB would most likely give him hell…and if they didn’t, we all know Fox News would. At least this time they would actually have a legitimate reason for their attack.

    Where you could take Milo in his adventure of life is an interesting thought though. How good or bad could you persuade him to be…it should be an interesting ride indeed.


  5. Geoff Keighley Says:

    Yep I did get to go hands (or “body”) on with it. It’s definitely very early on in development, but I’ll definitely be watching how it progresses.

  6. Rob "Azuma" Kimball Says:

    So was it as impressive as the footage made it look?
    How different was your experience from what they showed?

  7. chris Says:

    Wait a minute, is the real Dmitri Peter’s son or a relative? If not, what the hell is a young boy doing in Molyneux’s hotel room? That is creepy as hell.

  8. Matt Says:

    I’m proclaiming this worst game evr.

  9. TheEnd Says:

    They shouldn’t have shown that… filling the press conf. with a video of something that it’s not even runnable for a short live demo + a bunch of games that doesn’t come out until next year (what about the policy reg.not announcing games not coming in the next 6 months?) + talking about multiplatform games the rest of the conference only makes you wonder what will they be announcinge at Lepzing? or TGS?…Sony will be showing Heavy Rain, GT5, R&C for sure in Leipzing…Last Guardian, FF VS XIII in Japan…Microsoft?

  10. Joe P. Says:

    As others have pointed out, Geoff, it’d be nice if you could clarify a bit. Who is Dmitri, what was the young lad doing in the hotel room, how is Molyneux his role model, and why have the two spent so much time together over the years? As you’ve phrased this, it’s more than a little creepy.

    Secondly, I’m confused as to how long Molyneux has been working on Milo. 5 years or longer? On the one hand you say that it’s “been in development for years”, but at the same time it’s “definitely very early on in development”. That doesn’t seem to suggest that this technology will be living up to its already-ample hype any time soon, if ever. But then, some of us were already very skeptical about what’s been shown/promised so far, even without getting a first-hand glimpse.

  11. Daniel Says:

    Five years of development can be a short time, which everybody who works on a complex project knows. My PhD took 3 years, and it was on (relatively) simple stuff like image recognition.

    The first thing that came to my mind watching the demo was the console based psychiatrist someone wrote for Linux. This program was basically reflecting everything back at you (“Why are you saying this?”, “What do you think this means?”), but if you ignored this pretty blatant avoidance of really intelligent answers, you could have a reasonably complicated conversation with this shrink. Milo does the same thing in one or two instances, basically asking the player (Claire) to elaborate on something he apparently didn’t couldn’t make much sense of.

    The things Molyneux has added to the shrink program are basically (albeit impressive) new pieces of technology: Good movement recognition, image recognition and recognition of emotions in voices (although this seems to be tricky, which I think is why they restrict themselves to the more easily recognized ones such as “telling a joke”).

    The important point here is: True intelligence is the extraction of symbolic (or logical) information from literal information (images, sounds). Manipulating symbols in a given manner, for example checking a text for grammatical correctness, which newer text processors can do pretty well, is “easy”. Of course there’s lots of kinks to be ironed out before it works, but in the end it is just processing of symbolic information and checking it versus a table of allowed combinations.

    The main things Molyneux and his colleagues have been working on (apart from the simulation of Milo himself) all have to do with extracting information from raw data, and THIS is what allows true AI. If it works. So beware of MiloNet.

Leave a Comment