[00:39:52] *** Quits: Gurkenglas (Gurkenglas@dslb-178-000-176-040.178.000.pools.vodafone-ip.de) (Ping timeout: 244 seconds)
[00:41:16] *** Joins: Gurkenglas (Gurkenglas@dslb-094-223-143-032.094.223.pools.vodafone-ip.de)
[01:49:18] *** Quits: spectre (~fran@115.114.202.84.customer.cdi.no) (Ping timeout: 255 seconds)
[02:05:48] *** Quits: Gurkenglas (Gurkenglas@dslb-094-223-143-032.094.223.pools.vodafone-ip.de) (Ping timeout: 250 seconds)
[09:01:37] *** Joins: Gurkenglas (~Gurkengla@dslb-094-223-143-032.094.223.pools.vodafone-ip.de)
[09:10:17] *** Joins: spectre (~fran@115.114.202.84.customer.cdi.no)
[09:52:36] *** Quits: Gurkenglas (~Gurkengla@dslb-094-223-143-032.094.223.pools.vodafone-ip.de) (Ping timeout: 255 seconds)
[10:24:34] *** Joins: Gurkenglas (Gurkenglas@dslb-094-223-143-032.094.223.pools.vodafone-ip.de)
[10:33:04] *** Quits: drbean (~drbean@124.219.83.91) (Ping timeout: 265 seconds)
[10:37:34] *** Joins: drbean (~drbean@124.219.83.50)
[12:04:31] <inariksit> vin-ivar: yeah he is based in uppsala, he just came to gbg to give a talk in the meeting
[12:05:15] <inariksit> esg: it was fine, I didn't follow all the lectures, but there were many different topics and lots of stuff seemed interesting
[12:06:54] <inariksit> these dialogue people at GU have a funny corpus, they make people discuss in groups about a scenario where there is a balloon that is about to fall unless they throw away one person, and then they get description of people (doctor who says they are about to cure cancer ; pregnant woman, ...) and need to decide who to throw out
[12:14:38] <spectre> lol
[12:15:51] <spectre> they should get lesswrong involved
[12:43:32] <inariksit> :-D
[12:47:48] <spectre> there is probably a forum post where they can just c/p the thread into their dialogue system ;D
[12:51:30] <inariksit> hehe I think that what they investigate is related to it being a live dialogue
[12:51:34] <inariksit> but maybe they have irc logs
[12:51:44] <spectre> yeah...
[12:52:16] <inariksit> I asked who they usually throw out, and apparently some groups decide to throw everyone out
[12:52:21] <inariksit> not throwing anyone is not allowed
[12:52:24] <spectre> yeah
[12:52:31] <spectre> do any of them throw a random person out?
[12:52:36] <spectre> at my birthday we played a game
[12:52:51] <inariksit> but most of them don't have respect for the doctor, they reason that there's anyway a team of multiple scientists working on the problem
[12:52:58] <inariksit> and the doctor might just be bullshitting
[12:53:14] <spectre> called "vampire"
[12:53:46] <spectre> https://en.wikipedia.org/wiki/Mafia_%28party_game%29
[12:54:02] <spectre> i came up with the strategy of picking a person to kill at random
[12:54:17] <spectre> but then i randomed
[12:54:18] <spectre> and it was me
[12:54:23] <spectre> and they wouldn't let me kill myself
[12:54:24] <inariksit> :-D
[12:54:26] <inariksit> hahah
[12:54:30] <spectre> :/
[12:54:45] <spectre> have you played that game before ?
[12:54:46] <inariksit> I killed my partner in a game at choir weekend, out of desperation
[12:54:50] <inariksit> they were suspecting both of us
[12:55:01] <spectre> haha
[12:55:02] <inariksit> but yeah it was a stupid move
[12:55:56] <inariksit> btw I just made a SAT-free version of my CG and it performs better than SAT-version (expected) and also VISL-CG3
[12:56:02] <spectre> \o/
[12:56:04] <inariksit> well something like 30 words out of 22k
[12:56:13] <spectre> performs better than vislcg3 ?
[12:56:20] <inariksit> for this 22k spanish data
[12:56:25] <inariksit> I mean not speed but accuracy
[12:56:31] <spectre> ah
[12:56:32] <inariksit> and that is quite mystical
[12:56:33] <spectre> cool :)
[12:56:46] <spectre> yeah
[12:57:38] <inariksit> I just used my data type for rules, and most of the program is identical (how you interpret the conditions to sets to match etc.), but I just took the SAT part away
[12:57:47] <inariksit> and just run it in a loop rule by rule
[12:57:56] <spectre> could it be to do with the rule ordering ?
[12:58:07] <inariksit> aah yeah I ignored sections for this
[12:58:07] <spectre> is it identical in vislcg3/your program?
[12:58:28] <inariksit> ok so yeah disregard, it's most likely the section thing
[12:58:32] <inariksit> ok let's see
[12:58:55] <inariksit> anyway I'm getting a better result for 30 words out of those 22k, and that's the weird part
[12:59:01] <inariksit> even my recall is better than vislcg3
[12:59:06] <inariksit> well these are tiny things anyway
[13:00:13] <spectre> ah, that's a differential eval
[13:11:57] *** Quits: Gurkenglas (Gurkenglas@dslb-094-223-143-032.094.223.pools.vodafone-ip.de) (Ping timeout: 255 seconds)
[13:31:44] *** Quits: spectre (~fran@115.114.202.84.customer.cdi.no) (Ping timeout: 272 seconds)
[14:06:40] *** Joins: Gurkenglas (Gurkenglas@dslb-094-223-143-032.094.223.pools.vodafone-ip.de)
[14:10:04] <inariksit> after section, I get 18 different words
[14:10:05] <inariksit> still better
[14:10:09] <inariksit> so exciting :-D
[14:10:47] <inariksit> well this is the closest I've gotten to vislcg3 yet :') though I cannot call it sat-cg anymore
[14:11:46] <inariksit> (ok I'm not going to call this the finished product, just a sanity test whether the problems are in the sat-encoding or somewhere else)
[14:12:15] <inariksit> </monologue>, sorry about that :-P
[14:13:19] *** Joins: spectre (~fran@dhcp743-ans3.wifi.uit.no)
[15:57:59] *** Quits: Gurkenglas (Gurkenglas@dslb-094-223-143-032.094.223.pools.vodafone-ip.de) (Ping timeout: 244 seconds)
[16:15:09] *** Joins: Gurkenglas (Gurkenglas@dslb-094-223-143-032.094.223.pools.vodafone-ip.de)
[16:15:45] *** Quits: spectre (~fran@dhcp743-ans3.wifi.uit.no) (Ping timeout: 252 seconds)
[16:49:53] *** Joins: spectre (~fran@115.114.202.84.customer.cdi.no)
[23:08:37] *** Joins: Liscar (~canaima@190.205.11.72)
[23:08:49] *** Parts: Liscar (~canaima@190.205.11.72) ()