[MODULE] Megahal

ppslim · Post by **ppslim** » Thu Mar 20, 2003 11:04 am

It is likely you are using either a older version of windrop, that doesn't support customer channel settings, or a older version of megahal, that doesn't provide them.

sniiki · Post by **sniiki** » Thu Mar 20, 2003 11:29 am

Doh!
Tried never windrop (1.6.1 -> 1.6.13) and works fine!

Yep, it's a little stupid still, need tweaking.
But again thanks for the GREAT and FAST help ppslim!

bot quote:

[17:31:41] <@Sniiki> nutsylover: hi
[17:32:03] <+Nutsylover> Sniiki: Hi there. We're having some nice weather at the anus.

sniiki · Post by **sniiki** » Thu Mar 20, 2003 12:41 pm

Does this AI learn to speak better cos now it replys something about musical instruments when i say thanks?

ppslim · Post by **ppslim** » Thu Mar 20, 2003 2:18 pm

That is what AI is. Based on learned knowledge, it generates a reposce, that can changed based on what it knows about.

At some point megahal will stop learning new things to talk about. This is because the code that runs the module, starts thinking that the stuff it allready knows about, is better than learning more.

sniiki · Post by **sniiki** » Thu Mar 20, 2003 2:28 pm

how much/long is it able to learn (approximately)?

ppslim · Post by **ppslim** » Thu Mar 20, 2003 3:57 pm

It's usualy dependant on how broad you talk to it.

IE, if you ask it 5 or 6 things tiem and time again, or answer it the same things time and time again, it will run out fast.

If you ask it lots of questions, or change the structure of your sentances, it will last longer.

sniiki · Post by **sniiki** » Thu Mar 20, 2003 4:43 pm

ok thnx

DrAxman · Post by **DrAxman** » Tue Mar 25, 2003 8:04 pm

quick tip to get your bot to learn real quick create several eggys and run em from your machine each with a different brain file and maybe one using alice.tcl put em all in the same channel and start em talking to each other after 1 hour you have a massive brain file that is quite interesting and if as the previous person put you then put it into an abusive channel you get some interesting results

tainted · Post by **tainted** » Wed Mar 26, 2003 2:35 pm

I must advise anyone that is tempted to use this on a public shell to think twice. On a 200Mhz (hey, its a beast) freebsd box, with two people having a conversation with +learnall in the channel with the bot, it spikes to nearly 50% cpu load. On a 500Mhz fbsd shell, the cpu load for a small not-so-active channel with learnall took nearly 80% cpu load. Most shell providors will kill or even terminate people for huge cpu loads on their machines. I'm not trying to put this module down, it is amusing for sure, but its a bit buggy still it seems.

DrAxman · Post by **DrAxman** » Wed Mar 26, 2003 4:32 pm

you may notice i said run them on your own machine not shells also i found no probs running 6 bots on a 1200 athlon with 1 gig ram

egghelp/eggheads community

[MODULE] Megahal

Megahal learning

re learning