It would beat the liberals.

HAhahaha uber you're such a dingus.

^classic example of ruining and derailing every thread right here

I was trying to get a reaction from you bitches and it worked.

uh oh, he's trying.

ubercacher16 should find something better to do.

- Unfortunately not an official admin ruling

I do have better things to do, I just have so much more fun making DO mad.

I was talking to ghug about this last week. It seems like it should be possible to write a pretty good gunboat program, given that we have pretty good go and starcraft programs nowadays. The simultaneous moves make it a bit wonky, but should still be doable.

When you say black people are inferior, it makes Demon mad. When you do shit like this, it just reminds people how sad and pathetic you are.

There are gunboat programs out there. I don't think they're that great, but I couldn't speak to it personally.

Given the state of AI technology in general, I have little doubt that someone could build a diplomacy bot that could blow everyone away, especially in gunboat. Diplomacy is a little different than other problems in that you have numerous different actors rather than the two you have in games like chess, and in the simultaneousness, but at its core it's a pretty simple game.

Whether someone will ever put the time in is a tougher question.

But could you make a bot that could do diplomacy? That would be amazing.

A bot that could actually play the full game with negotiations would be much more challenging. I don't think the state of natural language processing is quite up to it yet. Generating press given a board state (including previous moves?), your power, the other power, and maybe the message history would be a pretty interesting task though.

http://www.ellought.demon.co.uk/dipai/daide_syntax.pdf

I don't think that counts as natural language.

A long time ago, I joined a judge game that required everybody to only write press in some machine-readable syntax. Unfortunately, there was no reference for this syntax, so every press I sent was rejected. That remains the only game I've ever CD'd.

It would be reasonably easy to make a bot that completely dominates in 1v1. The problem is that it might be a little predictable if you play against it multiple times-
- it might be good if you don't know that it's an AI but bad if you do. So Anon 1v1 is definitely the easiest problem. Gunboat would actually be a little tricky... having multiple players + hidden information makes Dip much more similar to Poker than Baduk or Chess. IIRC humans (especially computer-assisted humans) are still much better at Poker.

So the answer is no you can't just tweak a Chess bot to play Diplomacy well, you'd need to design something new.

I concur. AI playing against 6 at once would be really tough to be better than 1 for 7 or 2 for 7 overall. But, once near a midpoint (closing in on 1 v1) it might outperform.

Why are we disregarding Albert?

Alberts perform well against 6 humans!
Someone set up a host server (no deadlines) and keep it up, then we can try it out.

What's Albert?

Just googled it. Seems cool, but I think a neural net would likely be better than an expert system like Albert. I can't find any ML bots for Diplomacy.

I'm assuming an AI gunboat game would have different difficult to levels, but I wonder if when someone CDs, a bot could automatically take control (but perhaps the entire easiest level bot)?

I've played against Albert and it's not that great. I can solo against 6 Alberts about 1/3 of the time, so it's a nice tool, but as an AI it could use some work.

@JY, filling RT games was one of the main uses found for Daide's DipAI bots, though the developers were really competing to program the best bot. :)
It would be wonderful if Alberts could be used in Webdip as a game option, to fill in nmr'ers until a human takes over - though we now most of the hopeless posiitns are never taken, so it would be a great change to games suffering from giving up Cd'ers.

@LP, you must be a gunboat champion... winning 1/3 is great result!, have you realy tried the highest level, latest release?

@CCR My copy is at least ten years old. I didn't know they updated it.

And no I'm not a GB champ, I'm pretty average. My point was that an average player shouldn't be able to beat AIs 1/3 of the time.

The OP was asking in comparison to a chess program. For a casual chess player I'm good, but bad compared to club players. I can only beat most Chess programs at the lowest difficulty setting, so at least the version I have of Albert is nowhere near the sophistication of a chess program.