If you don’t know about Milo, it was a sweet tech demo for Project Natal (now Kinect) Peter Molyneux showed at E3 2009. There have been calls of ‘fake!’, and I’m sure the software for the demo was rigged to deal with specific behaviors. But I think Lionhead really was working on the demo/game (and hiring for it), even if it was subsequently cancelled and there were examples of smoke and mirrors being used in the demo. But, here, check out the original presentation in HD for yourself:
I’m presenting this doc just as an example of different directions you could take this kind of game in; as an example of an extended prototype/pitch document; and just because when I re-read it, I still liked some of the ideas! What you think? — Would you enjoy playing one of these scenarios? Would your kids? (Or cousins, nephews/nieces, or whatever.)
All rights to the following pitch document, along with the ideas contained within it, have belonged to me insofar as legally possible since 2009. Without further ado, here’s the pitch!
Benefits and aspects of interacting with Milo through Project Natal:
- The sense of an ‘imaginary friend’; emotional AI
- The illusion of intelligence through carefully managed interactions
- Control in the game world though the player’s movement in their own real space
Game Genre; and Requirements in being a Game:
Milo has the potential to create a whole new game genre. Closest existing game genres to the Milo tech demo’s current characteristics:
- Tamagotchi or Petz-style ‘take care of him’ games
- Proofs-of-concept for ‘exploring a world’, e.g. HL2: ‘Lost Coast’ demo
What the Milo tech demo is lacking which games require:
- Competition, challenges and rewards, balanced by risk
- Beginning, middle and end
- A clear sense of what Milo needs from you, what your objectives are as a player
Three Options for Milo’s Future
Primarily a limited conversational AI, with whom you explore a self-contained, small world with ‘minigame’-style challenges. Essentially, you log in to ‘hang out’.
We’ll call this: FRIEND SIMULATOR. This is the closest leap from what we’ve seen of Milo’s world so far.
A boy whom you ‘meet’ at the ‘beginning’. You form a basic relationship with him. He has tasks around the house with which you help him. These could either be periodic/episodic – i.e. a different one each time you play, self-contained (easier for the AI’s memory code) – or structured into a basic progression of difficulty, with Milo remembering your previous assistance as new issues arise. For instance, in a science project, you accidentally pollute the pond. Then later on, Milo’s mother makes you both clean it up! This is more complex for the AI’s memory: in order to be really convincing, Milo should remember the specifics of what you did earlier (not just any player, but specifics that were identifiable to you only). Once all tasks are complete – whether the game is organised episodically or in an ongoing progress – the game is ‘finished’ and Milo is a ‘friend’ whom you log on just to ‘see’ (i.e. it comes to resemble game type A, after completion of the game).
We’ll call type B: SIMPLE STRUCTURED GAME.
Again, a beginning and an end. Again, a conversational AI that persists indefinitely as a ‘friend’. But in type C, the game is driven as a goal-oriented experience by a full STORY, with developments (including changes in character), risks, rewards, waypoints, enemies/bosses/challenges, and an overarching plot that ties it all tightly together.
We’ll call this a: COMPLEX STORY-ORIENTED GAME.
Developing Concept C
Each of the three types – A, B, and C – is both feasible and interesting in its own right. They also fit different budgets, and expectations will need to be managed differently according to the game type. But, to me, C is the most exciting, and I will expand on its possibilities now.
RISK: As in any game, introducing a story creates risk for the developer. It must be strong and appropriate. In a game which (so far) has not suggested such depth, advertising would need to expand upon the scope and nature of the game.
DEEPER EMOTIONS: The character development in Milo, and in the player, that accompanies a full story can create real emotional investment, and its challenges define the purpose and experience of the game in ways that are both more magical and also more saleable than a simple ‘proof-of-concept’-type experience.
AWARENESS: As we saw to some degree in the tech demo (‘Don’t forget your Mum’s birthday!’), with AI comes the possibility of some self-awareness. However, it is a great mistake to conclude that this limits the options as far as a fantastical story is concerned. This should be viewed as a very exciting chance to have a self-aware fictional character willingly and openly draw the player into his fantasy world. By breaking the fourth wall, we to some degree remove the player’s/viewer’s sense of distance and condescension. If Milo is aware of his limitations as well as the power of his fantasy world, then he knows as much as we do, and this enables him to become a truer and more meaningful person.
Furthermore, by approaching the character this way, we nullify the ‘realism disconnect’ in situations which are too complex for the AI’s capabilities. For example, if Milo can frankly say, “I really don’t understand what you’re talking about now: remember, I’m running on a games console, I’m not Skynet or anything…” he can defuse and eliminate some of the traditional and irksome problems with AI as seen in earlier games where, for instance, AI ‘co-op’ is supported. Obviously, the better the underlying AI, the less frequently this self-deprecation will be required. The bottom line is, breaking the ‘illusion’ of a real person might turn out to be the best way to make Milo even more effectively real.
Matters of Story
Stories require two things to work.
Otherness: The story cannot be ‘just another day’. Something has to be special – something unusual occurs. In film plots this is called the ‘inciting incident’ – it sets our hero on his path. I put forward the suggestion that Milo’s world starts out very similar to our own, and that some change or revelation in this world kicks off our story.
Here are some possible examples of inciting incidents:
i) You and Milo discover a gateway to a strange world together, and you cross through.
ii) Your arrival in Milo’s world is the inciting incident: he becomes self-aware as an AI, and realises that his whole world is malleable. His curiosity drives him to experiment with the world.
iii) When Milo meets you, he decides that his life is dull, and the rules of the computer game are unfair. Together you try to break free from the forces of computer game architecture, overcoming the limitations of its programming. Perhaps Milo’s goal is to cross over into the real world? During the course of the story, Milo could influence your real-world life from within the game. At the end, he could ‘break free’ – and then he would only return to the game world if you ‘called him back’.
Conflict: The other thing all stories need is conflict, whether it be emotional, ideological, physical, or (most likely) a combination. Our examples of inciting incidents have already given a sense of this:
i) In the magical world there is an evil force to be vanquished.
ii) Milo is in conflict with his self-aware boredom, which leads him to create new artificial parts of his world. He loses control over these, and then he needs your help to escape, master, and possibly reverse these changes.
iii) Milo becomes ideologically discontent with being an AI trapped in a rigid game, and tries to overcome the philosophical limitations of this existence.
As you can see, all of the example story premises have a form of conflict within them – nicely split, coincidentally, into physical, emotional and ideological jeopardies.
Conflict, in any form, can probably be summed up neatly:
MILO HAS A PROBLEM.
The unusual thing about this story setting is that ideally both Milo and the players are the heroes in this game. This is what having a player and a full AI in combination should mean: it’s multiplayer, but with only one human.
A Simple, Local, Story-Based Game
To go back to a simpler example for the moment, we should remember that Milo’s ‘problem’ can be quite straightforward and restricted while still creating a good game. Here is an example of that.
Milo’s parents had to go on an emergency business trip to protect their ownership of their farm. They left Milo some instructions for how to take care of things, but matters are getting a little out of hand – it seems that the farm has some very unusual qualities! Help Milo set things to rights before his parents get back!
In this example Milo’s problem is twofold: he has to meet the expectations of his parents; but also, the unusual surprises in the farm present challenges. For instance, all the water for irrigating the fields runs through Mister Thomas’s land to get to Milo’s farm. Mister Thomas has cut off the supply – and he’ll only restore it if you complete three seemingly-impossible tasks for him.
So, challenges and some drama certainly can be constructed within a limited framework. However, it is important to remember that the amount of drama here, and hence the degree to which Milo and you can develop as characters within this world, is limited.
Milo Alone as a First Instalment
A development of the Milo Alone concept would be to make it a ‘first instalment’, or episode, of an ongoing Milo series. This has benefits in production, budget, and marketing. For instance, Milo Alone could introduce you to Milo and his basic world. It would test the terminology and be contained enough for Lionhead to release sooner rather than later. It could present interesting sequel opportunities, if Milo’s parents promise, if he succeeds in his tasks, to show Milo the farm’s ‘secret’. That might be the gateway to another world; or an amazing tool that lets you and Milo draw things that appear in the world; or any of a host of other possibilities that might best be reserved for a second game with more mature technology and an audience that has already learnt the ropes.
Speaking of code, a miscellaneous idea for improving interactions with Milo:
Assuming that Milo’s frame of mind, emotional state and maybe even vocabulary can be weighted statistically, the game could ask a new player to give their age. This could be used to modify to some extent how Milo behaves with them. This could ensure that, to some degree, the ‘Jar Jar Binks effect’ can be avoided. When playing with adults, Milo is a bit less bouncy, a bit more ironic, a bit toned-down. When playing with little kids, Milo is just as jubilant, carefree, thoughtless and impulsive as them.
Milo could occasionally ask the player to bring along a friend to help out with a problem that simply requires more than two people. This could introduce not exactly a ‘multiplayer’, but definitely another human, co-operative element within parts of the game. The dynamic between two humans and Milo will also be different: and Milo’s different relationships with the players will make his interactions with each of them unique; as will the relationship between the two humans. When bringing a friend to play with Milo, you might start to feel as though you were introducing two friends who had not met before. The social dimension of this situation is fascinating and could broaden and intensify the game – as well as reduce the possible notion in the media that a virtual playmate is a destructive social ill, rather than an opportunity to experiment with new forms of socialisation.
Will Milo be able to voice- and face-identify unique players, making our real connections with him stronger, and avoiding masquerading? Obviously, it pretty much defeats the point of becoming his friend, if anyone can log in as you, say your name, and Milo will treat you as if he has an extended relationship with you.
What is the extent of Milo’s ability to remember facts, occurrences, and people? I’m assuming that there is, or is under development, some kind of database system for storing relevant ‘memories’ and information.
If, for instance, Milo Alone were to be a first instalment, would it be possible to create a provision for maintaining your existing relationship with him when starting the next instalment? This could be fairly important.
Food For Thought
As you can tell, I’ve been thinking about Milo a fair bit. I would be delighted to come to Lionhead and discuss my ideas, and your ideas, about the possibilities for Milo. The implications are extraordinary, and it would be a privilege to join your team in the attempt to usher in a new era for games and for culture more widely.
Hope you enjoyed that. In 2010, a year after I wrote the document above and sent it to Lionhead via their recruiting service, Molyneux did a follow-up talk at TED. See if you think they incorporated any of my notes 🙂