Experiments in Synthetic Virtual Worlds?

In the late 1980's I read Valentino Braitenberg's Vehicles: Experiments in Synthetic Psychology.  I was charmed by the thought experiment; I was seduced by how one might create intelligence by layering the building blocks and observing the evolution of behavior within an environment...  Last week I revisited Vehicles.  I thought about it again, with a different viewpoint: was there a useful idea here for virtual worlds research and discussion?

Cosma Shalizi provides a good review of Vehicles.  The secret of the seduction, is Cosma's point:

(the) law of ``uphill analysis and downhill synthesis'' applies... it's easier to design a mechanism from scratch to do something, than to figure out just how nature has contrived to do it; this suggests that maybe the natural way isn't really insuperably complicated.

The table of contents suggests Braitenberg's synthetic progression for Vehicles:

Vehicle 1. Getting Around
Vehicle 2. Fear and Agression
Vehicle 3. Love
Vehicle 4. Values and Special Tastes
Vehicle 5. Logic
Vehicle 6. Selection, the Impersonal Engineer
Vehicle 7. Concepts
Vehicle 8. Space, Things, and Movements
Vehicle 9. Shapes
Vehicle 10. Getting Ideas
Vehicle 11. Rules and Regularities
Vehicle 12. Trains of Thought
Vehicle 13. Foresight
Vehicle 14. Egotism and Optimism

A while back I was chatting with someone about games, a point made in passing stuck with me.  It went something like this (a liberal reinvention from memory):  "the frustrating thing about talking about games research is that it  quickly becomes anecdotal..."   

Perhaps a similar point can be raised about virtual game worlds analysis/discussion: they seem to be so contextualized in terms of game design (choices) that they become hard to communicate without a lot of explaining.  E.g. try explaining the implications of permadeath in game design.

My question is, do you believe this?  If so, in the spirit of what Braitenberg's thought experiment for cognitive science, is there too a "downhill synthesis" thought experiment, waiting,  for virtual worlds and how many chapters would it take?

Ahem,

Virtual World 1.   A Simple Undulating Landscape, Empty and Without Wind...

Comments

[Edward Castronova]

1.

2. Resources (digital objects with varying characteristics) are distributed across the landscape.

3. A technology is imposed. (technology = rules allowing resources to be manipulated, changed, and combined, with knowable results.)

4. An interface place the technology and items within the control of users.

[Richard Bartle]

2.

Nathan>Virtual World 1. A Simple Undulating Landscape, Empty and Without Wind...

Don't you mean "Let there be light"?

Richard

[Flatfingers]

3.

I remember bumping into Braitenberg's Vehicles concept nigh onto 20 years ago, and getting so excited about exploring the "emergent behavior" effect that I sat down and designed a Vehicle Structure Language.

What I got out of the concept wasn't so much "here's how to create intelligence," but more "wow, it sure is easy to fool humans into thinking that some built system displays intelligence." The Vehicles thought experiment seemed to say more about human perception than about engineering.

With respect to the forms and functions of objects and processes found in nature, its advantage over human goal-directed design is having a sufficiently long period in which to throw away the myriad low-payoff designs. For a real eye-opener on how humans err in trying to short-circuit this design process, I've never found a more enlightening guide than John Gall's _Systemantics_. (I use the word "enlightening" deliberately; there's a certain Zen-like quality to Gall's lists of Horrible Examples of human design.) The system design principles in _Systemantics_ should be part of the toolkit of anyone with system design responsibilities, and it makes for pretty interesting reading from a natural scientist POV as well.

Which brings me to virtual worlds. It's probably obvious, but put me down for their development being incremental over time, rather than springing full-formed from some Zeus' brow.

I'm actually optimistic about this; I think we're approaching a very exciting time in VW design. With continuing boosts to personal computer processing power, connection speeds, and overall connectivity, I suspect we're heading toward a period of "garage VWs." People will have the tools to build experimental but still attractive VWs, free of commercial requirements or creative restrictions. (With respect, I don't believe the handful of large corporate VW designers, with their financial responsibilities, are able to offer this level of "try it and see if it works.")

Sure, most people will never try their hand at VW design even if they do have the computing horsepower and the tools. And sure, 90% (at least) of the homebrew VWs that are built will be crap, or to put it more politely will have only very limited appeal. But the few that are genuinely inventive will seed the next generation of VWs with the concepts they pioneer. In effect, I believe we're approaching at a time of fertility that, like nature's, will be effective because it can afford to try enough different things to throw away what doesn't work.

[Brian Yeung]

4.

What I got out of the concept wasn't so much "here's how to create intelligence," but more "wow, it sure is easy to fool humans into thinking that some built system displays intelligence." The Vehicles thought experiment seemed to say more about human perception than about engineering.

Isn't a lot of design about creating not necessarily a "realistic" world, but rather a convincing and consistent one? Or in theater terms, having the highest production value for the lowest development cost?

[Flatfingers]

5.

Brian Yeung: "Isn't a lot of design about creating not necessarily a 'realistic' world, but rather a convincing and consistent one? Or in theater terms, having the highest production value for the lowest development cost?"

As I keep trying to explain to one of the people on my team who's been given some design responsibility, scientists and engineers look at the world in at least one crucially different way: where the scientist expects perfection, for the engineer "good enough" (within reasonably tolerances) is acceptable.

This difference starts to become a factor when these folks get put into a design role, particularly of human systems. The engineer is inclined to leave systems a bit slippery, which often winds up being acceptable in systems design but can sometimes lead to mild sloppiness ("if there are any problems the software guys can fix them later"). The scientist, on the other hand, tends to overdesign in the belief that only perfection is acceptable, and sometimes winds up spending a lot of time polishing some interesting but small function.

All of which is to say that while I agree with you in theory about the value of evaluating designs on their likely payoff (that is, most benefit for least cost), I've found that principle very difficult to put into practice because people by nature have very different understandings of "value" and "cost." A VW designed by a scientist would look quite different from one designed by an engineer because their perceptions of what rules make for "good" design are very different.

This is one reason I'm looking forward to more "garage" VWs -- the more of these things we can build from more unique perspectives, the more likely we are to develop the new paradigms that Richard first mentioned.

[Nathan Combs]

6.

2. Resources (digital objects with varying characteristics) are distributed across the landscape.

3. A technology is imposed. (technology = rules allowing resources to be manipulated, changed, and combined, with knowable results.)

4. An interface place the technology and items within the control of users.

Regarding the player side of the equation. What should be delineated into these chapters: the player, or the community?

One view might be to vary the society machine (neatly typed by classes of NPCs) and view the player through the lens of differing community and world assumptions.

[Nathan Combs]

7.

What I got out of the concept wasn't so much "here's how to create intelligence," but more "wow, it sure is easy to fool humans into thinking that some built system displays intelligence."

Or in theater terms, having the highest production value for the lowest development cost?

So, we are a cheap date?

Don't you mean "Let there be light"?

"Boom! Boom!" (Basil Brush)

[Richard Bartle]

8.

>"Boom! Boom!" (Basil Brush)

You spent too long in England, Nate!

Richard

[Richard Bartle]

9.

Nathan Combs>In the late 1980's I read Valentino Braitenberg's Vehicles: Experiments in Synthetic Psychology.

I finally remembered the game this reminded me of: Mind Rover.

Richard

[Brian Yeung]

10.

NC: So, we are a cheap date?

There's a good discussion I missed...

I suppose my comments should be given the caveat that they're coming from someone new to designing VWs, and still learning what's overdesign and what's "good enough," given looming project deadlines :)

[Hellinar]

11.

Flatfingers >I'm actually optimistic about this; I think we're approaching a very exciting time in VW design. With continuing boosts to personal computer processing power, connection speeds, and overall connectivity, I suspect we're heading toward a period of "garage VWs." People will have the tools to build experimental but still attractive VWs, free of commercial requirements or creative restrictions.<

As someone who has been putting together a "garage" VW for the last four years, I can attest to the increasing power of the tools. I feel vastly more productive this year than last, with the improvements in the tools I am using. (Java IDE and OpenGL interface). And the increasing speed of machines makes new approaches to emergent worlds possible.

I my case, I am setting up a VW to support a gardening game. So my main focus is on procedural methods of plant growth. To date, far as I am aware, digital plants have been based on procedurally assembling varied components. Leaves, branches, blooms, bark etc. With some success, as in the SpeedTree system. But increasing processor speed has let me use growth based on functionally identical cells controlled by genetic switches. This method has been well proved by evolution, and I'm very happy with the results. I think we will see a lot more tipping points like that, where emergent methods become practical just from sheer increase in tool sophistication and processor power.

Though I might have been able to imagine the same code five years ago, I personally couldn't have wrote it. I'm a fairly careless programmer. Without powerful debugging tools to track down the many unexpected side effects of emergent code, I would long ago have given up in frustration.

[Nathan Combs]

12.

Richard:

I finally remembered the game this reminded me of: Mind Rover.

The PS2 game, Magic Pengel presents a fascinating means of synthesizing characters. I've never played it, but apparently one can "doodle" warriors:

the most important tool in the entire experience has to be the Paint Shop inspired illustrator that's used to draw your assistants. Beginning with only four colors, one brush type, and two body variables, your early creations aren't going to be much. As your skills and experience improve, however, the things you can do with the paint program become enormously elaborate. And while your only limitation is the confines of a multi-speed cursor and a total of 64 body parts, skilled artists will be able to create some seriously beautiful images.

As an added bit of strategy, the types of colors in use, thickness of your brushes, and placement of certain body parts affect the power and attributes of your drawings. Use a lot of red, for instance, and your character's hit points increase. Add a touch of blue, and your luck goes up a few notches. There's a full rainbow of colors to experiment with, and the number of limbs, pieces, and other such factors all make a difference in your creation, with a potentially unlimited amount of possibilities. Artists are going to love it.
(from here)

[Philip Rosedale]

13.

Required reading, that book! We've got it floating around the office here at Linden Lab. Very fun way of explaining ideas around emergent complexity in simple systems and cognition.