He wrote a long post, but here is the beginning:
I want to see procedural content not just create the raw terrain maps, but thousands of tribes, nations and empires. I want it to generate language, culture, aesthetics, architecture, religion, scientific progress, humanity. I believe that many different game engines should connect to a single cloud data application in order to create persistent worlds. Strategy gaming should prepare for next generation graphics technology, neural network AI, and implement many different portals to access game information. Taking exponential trends into consideration, I want to see what a strategy game looks like when millions of semi-intelligent “agents” compete or collaborate for resources.
I’m talking about Guns, Germs and Steel in gaming form. It could answer questions about human settlement patterns. With different continental configurations, do certain types of regions always become colonial powers or is having many states in feudal competition becoming market powers all that is needed? Does this usually follow the parallel latitude crop theory? The game could have an arcade mode like the Civilization and an observer mode: set the stage and watch. Go back in time, change a few variables and watch the difference. Maybe I’m alone, but that type of concept excites me!
So one of the first thing’s that came to mind was Palantir’s software:
And taking their idea of reducing friction between human and computer to enhance human capabilities:http://www.palantir.com/2010/03/friction-in-human-computer-symbiosis-kasparov-on-chess/
Also, here’s an overview of some of the older high-dollar ABM projects that were implemented in the past, mostly using JAVA: http://home.comcast.net/~dshartley3/DIMEPMESIIGroup/ModelingSimulation.htm
Combat will change over time like you said. One of the models being used in real life is generational, 1st to 4th generation warfare (see: John Robb, Global Guerrillas), with theoritical 5th generation warfare. So you would have to choose how realistic or metaphorical you would render that action. You know, if you had a civilization event where spearmen destroy a tank, or if you could zoom straight down and play 1st person as the spearman trying to exploit the terrain, things like that. Then you move into modern things like winning hearts and minds, then into the vague world of secrecy and influencing complex systems in 5GW.
Economics will change over time.
If you’re creating language, then you’re also creating the box within which people think to some degree, so that creates a feedback loop with culture. So you have a have to go a Systems Dynamics approach with stacks, flows and feedback loops, and possibly going much further because you have to extend the model so far out.
You will have to measure emotional reactions as well, but the question is how. Are you going to use the traditional valence and arousal model, to model how a given population is reacting to stimulus? I’m thinking back to OpenCog’s agent and it’s ability to become “scared”, or inquisitive because it uses an economic attention allocation model.
I don’t think Jared Diamond’s theory is going to cut it for explaining everything to the degree of simulating this world though from a historical perspective either. Diverse agents who are connected with interdependent relationships that adapt over time.
Bleh, that’s all I got right now.
Also, if you could do that, it might be realistic and immersive enough that people would pay you to test and develop it, like a subscription based MMORPG-ish thing.
See also: Game Mechanics: Advanced Game Design by Ernest Adams and Joris Dormans
It’s on amazon and “other” sites.
If you wanted to introduce more variety into the game, instead of following a fixed technology tree, perhaps it could implement different concepts of the singularity towards “end game”. You could shift currency over time, as processes become automated you would move from a coin based economy to a paper one, then from paper to digital currency, then possibly to currency based off of:
- Energy – This may or may not be relevant, on the one hand advanced solar panels and cells, combined with nuclear power makes energy abundant. It remains to be seen how much power future computing will eat up. Current power drain on large systems comes not only from the computer, but the costs of cooling. Intel has been working on low power mobile processors:http://newsroom.intel.com/community/intel_newsroom/blog/2012/09/11/intel-low-power-processors-to-fuel-future-of-mobile-computing-innovation
- Antimatter: http://en.wikipedia.org/wiki/Antimatter#Cost Scientists claim that antimatter is the costliest material to make. In 2006, Gerald Smith estimated $250 million could produce 10 milligrams of positrons (equivalent to $25 billion per gram); in 1999, NASA gave a figure of $62.5 trillion per gram of antihydrogen. This is because production is difficult (only very few antiprotons are produced in reactions in particle accelerators), and because there is higher demand for other uses of particle accelerators. According to CERN, it has cost a few hundred million Swiss Francs to produce about 1 billionth of a gram (the amount used so far for particle/antiparticle collisions). Several NASA Institute for Advanced Concepts-funded studies are exploring whether it might be possible to use magnetic scoops to collect the antimatter that occurs naturally in the Van Allen belt of the Earth, and ultimately, the belts of gas giants, like Jupiter, hopefully at a lower cost per gram.
- If computers take over most of the operations in a society, costs can be based off of CPU cycles
This raises the question of whether space exploration would be useful in this context. Bill Stone has an old but good TED talk on space exploration and his on-going work to journey to the moon and his plan to mine the fuel to return home from the moon itself:
In the computers rule everything dynamic, Hugo de Garis has some interesting ideas on “Artilects”:http://profhugodegaris.wordpress.com/artilect-polls/
One field no one mentions: Hardware security. Trying to beat computers through software is nice and all, but there are many hardware bugs, BIOS rootkits (technically software), using firewire to pass through security, USB sticks with malware and keyloggers (which can be built into a keyboard). Unless an AI/AGI has built in defense to make the hardware difficult to get at, to self destruct, or reprise against attackers, then they are a vulnerable target.
Eclipse Phase also has interesting implementations.
I like everything about what you just said. I’m familiar with the Van Allen Belt antimatter and Hugo de Garis but I wasn’t familiar with Bill Stone, that’s pretty awesome! And the Adams-Dormans book looks awesome! Currently I’m reading this book my friend lent me. (Glad I didn’t have to pay for it!)
I am a big proponent of a genetic algorithm based tech tree that builds momentum towards a singularity end game. Not sure how it could be implemented at first. In an ideal world the engine is sufficiently advanced to model physics and the agents experiment with the physics engine. For the short term, I think breaking all technologies into their basic components based on physics would give a good “lego set” for building technologies. Each component could have a weight. Every time it gets used in a technology it gets a stronger weight. Etc. The momentum builds.
So you think hardware security makes AGI vulnerable? I suppose it’s an engineering question. Right now it’s vulnerable, but so were the worlds first single cell organisms. I’ll bet security increases over time as neuromorphic chips become more complex; maybe not though.