Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Shonai_Dweller]

I feel like you could just email this thread to Google and one of them would get on it, either for fun or as an official project.  Kind of the same way I think that if we sat down and showed this game to JJ Abrams he would immediately get started on Dwarf Fortress the movie.

But, how could you judge how well it's learning? It could spend the first few years generating the perfect world and tweaking raws (like some players surely do...) without ever starting up the game. Is that success?

How about "Learn to play the game and be able to report on it's own adventures by being able to judge what the community would think is a pretty cool story." Now that's a challenge I'd like to see an AI attempt one day. Much more impressive than playing Go by itself a billion times.

That sounds far, far beyond our current understanding of artificial intelligence, js.

Well, you wouldn't need Google to invest billions in working out the solution if we already understood it.
Go for it Google, lets see your ai play DF a billion times and pick which story would enter the Bay 12 Hall of Legends.
You're allowed to post twice a week on What's happening in your fortress, to gauge reaction.

[Vuohijumala]

Well, you wouldn't need Google to invest billions in working out the solution if we already understood it.
Go for it Google, lets see your ai play DF a billion times and pick which story would enter the Bay 12 Hall of Legends.
You're allowed to post twice a week on What's happening in your fortress, to gauge reaction.

Hey.. that sounds like a great idea!

Seriously, I'd sign a petition for this.

[xominxac]

Well, you wouldn't need Google to invest billions in working out the solution if we already understood it.
Go for it Google, lets see your ai play DF a billion times and pick which story would enter the Bay 12 Hall of Legends.
You're allowed to post twice a week on What's happening in your fortress, to gauge reaction.

Hey.. that sounds like a great idea!

Seriously, I'd sign a petition for this.

Hmm... Other than someone close to Google sending them an email, Change.org?

[NW_Kohaku]

In all seriousness, this is actually one of the worst types of games to give to a learning AI. 

What counts as success is highly subjective in DF.  You're asking for an AI not to achieve some objective win condition but to make "the most epic game story".  An AI learns a game through measuring whether it achieves objective goals more efficiently through one strategy or another.  It would first have to develop the current Bay12 concept of what is "epic", and then have to learn that doing a slight variation of the thing it already did will "feel less epic" even when it actually achieves incrementally more significant results, just because it is less novel to the audience the second time.

[Shonai_Dweller]

In all seriousness, this is actually one of the worst types of games to give to a learning AI. 

What counts as success is highly subjective in DF.  You're asking for an AI not to achieve some objective win condition but to make "the most epic game story".  An AI learns a game through measuring whether it achieves objective goals more efficiently through one strategy or another.  It would first have to develop the current Bay12 concept of what is "epic", and then have to learn that doing a slight variation of the thing it already did will "feel less epic" even when it actually achieves incrementally more significant results, just because it is less novel to the audience the second time.

Exactly, it'd have to learn about human (as far as that applies to Bay 12 members) interaction. That's far more impressive and worth throwing money at than playing chess by yourself a billion times. You've gotta think big if you ever want to reach Terminator levels of awareness.

DF has the added bonus of forcing the ai to deal with bugs from the variety of options out there. Real-world problem solving ai. Sounds good. Call Google.

[AceSV]

That would make the AI all the more interesting.  The Go-AI is based on something similar to a chat-bot, studying established patterns and mimicking them, so it actually should try to make fortresses that are as epic as possible. 

Teaching an AI to achieve an abstract victory would be an amazing step forward in the field of AI.  I mean, that's art, exploration, invention, the real human essence. 

[NW_Kohaku]

That would make the AI all the more interesting.  The Go-AI is based on something similar to a chat-bot, studying established patterns and mimicking them, so it actually should try to make fortresses that are as epic as possible. 

Teaching an AI to achieve an abstract victory would be an amazing step forward in the field of AI.  I mean, that's art, exploration, invention, the real human essence.

That depends mightily upon how the AI actually measures "victory", however.  It can't form it's own method of judging its own playstyle without first becoming advanced enough through human feedback, so it needs constant human feedback. 

Unless it's more advanced than most chatbots, it's also not necessarily capable of understanding positive from disinterested from negative feedback or even people arguing with one another rather than arguing with the AI.  If you are capable of making a chatbot THAT advanced, you might as well release the chatbot without making it play DF, because that's news in and of itself, and it would probably be able to convincingly fake playing DF, anyway, just by reading enough forum posts and regurgitating a convincing arc of competency gain. (Oh God, how many of them are already among us?!)

If you have to feed it objective feedback, you're pretty much stuck with giving people one of those dials that goes from 1 to 10 they give to people watching debates to see how much they liked individual parts of the speeches, and having them somehow sit and watch thousands of hours of gameplay while in the same state of mind for judging someone else's game...