I'm not exactly sure how you make the cutoff. Let's say that you are writing a bot and you want to be able to allow it to say, "Fa is free beer". How are you going to do this? You are going to keep track of how much phaser damage you inflicted on him plus how much torp damage you inflicted. Let's say you are a stickler for details. You then also calculate how much damage he took from torps that were detted plus torps that hit nearby teammates. You then also keep track of his repair rates. Finally, you include damage for planets that he's flying nearby. You have it worked out so that you can estimate within 5 points of where his shields and hull are. Now, when you are done with the dogfight, what are you going to call out? "Fa is free beer" or "Fa has 0% shields, 31% hull, and 22% fuel"? Let's say that the bot was playing by itself with other human teammates. Which would you rather hear? Hell, let's say that you designed some macros to send this information out. A good player can probably estimate with 20 points what the person's combined shields and hull are at and the fuel can probably be estimated within 1000. If I had macros to send out info like "Fa is without shields and under 25% hull and 25% fuel" would that be cheating? What if I was a super speed typist and got very good at estimating a person's damage? Would I not be able to communicate this info with my team? What if I was in a room with a teammate? Would I have to type this over the message board or could I just load my speech program (sarcasm alert) and "speak" the information and hope that his "voice recognition" program could properly decipher my meaning? I think that applying human ways of doing this to how bots should work is somewhat unfair and pointless. Do we feel that computers cheat at chess because they just look many moves ahead using a brute-force alpha-beta pruning technique rather than actually analyze a position? Now, in practicality, I agree with you on many points. One, a bot has to be smart enough to do the right things even with playing with human teammates. Why? Well, because I had some generous server gods who let me put FreeBot on their server to do some testing. If I required teammates to open up a socket connection to communicate with the bot, I think that would've come to a halt fairly quickly. Personally, I'm intrigued by how a bot performs with human teammates. If it can do that well, having it perform with fellow bot teammates shouldn't be much more difficult. Brian ----- Original Message ----- From: "Dave Ahn" <ahn at vec.wfubmc.edu> To: <vanilla-list at us.netrek.org> Sent: Tuesday, December 11, 2001 4:25 PM Subject: Re: [Vanilla List] organized, intelligent 'bots > On Tue, Dec 11, 2001 at 12:41:53PM -0500, Mark Mielke wrote: > > > > Heck, all of ICQ / MSN / AIM use "sockets" to communicate. Is it illegal > > for a human player to make use of any of these mechanisms of communication? > > There is a difference between improving the interpretation of the same > information that is available to every player and gaining new information > that is not supposed to be available to every player. The more you rely > on the latter to build your robot team, the less credible and interesting > your experiment becomes. > > Consider this example. Human players often call "Fa is free beer" to > indicate a damaged ship. Suppose that your robot team kept a global and > up-to-date list of all ship stats that are shared by all robots and used it > to guide the team strategy. Would it work? Probably. Is it "fair?" > Probably not. Is it impressive? No. > > An important part of the netrek game is the ambiguity and the unknown. > Entire games are won on the basis of a fake warp-1 SB that draws oggs at > the end of regulation while teammates drop for the win. If you have to > rely on GOD-MODE to make up for the insufficient AI heuristics in your > robots who are unable to function under the same conditions as humans, > then your project is simply a math exercise. > > If I were writing an AI robot, I would modify the Vanilla code to allow > observers to pilot a ship and get it installed on a public server. I > would then create an AI robot that could fly the ship to be at the > right place at the right time and send a message to the team saying, > "I should be here, doing this." If the robot were correct at least 50% > of the time, it'd be very impressive. > > -- > Dave Ahn | ahn at vec.wfubmc.edu | Wake Forest University Baptist Medical Center > > When you were born, you cried and the world rejoiced. Try to live your life > so that when you die, you will rejoice and the world will cry. -1/2 jj^2 > _______________________________________________ > vanilla-list mailing list > vanilla-list at us.netrek.org > https://mailman.real-time.com/mailman/listinfo/vanilla-list >