| --- Log opened Sat Jan 09 00:01:00 2016 | ||
| -!- jtimon [~quassel@103.red-80-26-235.adsl.dynamic.ccgg.telefonica.net] has quit [Ping timeout: 272 seconds] | 00:03 | |
| -!- pompolic [~A@unaffiliated/pompolic] has quit [Ping timeout: 276 seconds] | 00:41 | |
| -!- FourFire [FourFire@cm-84.215.195.59.getinternet.no] has joined ##hplusroadmap | 00:45 | |
| -!- pompolic [~A@unaffiliated/pompolic] has joined ##hplusroadmap | 00:49 | |
| -!- pink_vampire [~pink_vamp@67.210.40.189] has quit [Read error: Connection reset by peer] | 00:50 | |
| -!- pink_vampire [~pink_vamp@67.210.40.189] has joined ##hplusroadmap | 00:51 | |
| -!- jenelizabeth_ [~jenelizab@cpc76802-brmb10-2-0-cust399.1-3.cable.virginm.net] has quit [Ping timeout: 240 seconds] | 00:56 | |
| -!- jenelizabeth_ [~jenelizab@cpc76802-brmb10-2-0-cust399.1-3.cable.virginm.net] has joined ##hplusroadmap | 01:02 | |
| -!- NixCipher [~NixCipher@96.44.145.226] has quit [Ping timeout: 264 seconds] | 01:04 | |
| -!- sandeepkr [~sandeepkr@111.235.64.4] has joined ##hplusroadmap | 01:15 | |
| -!- NixCipher [~NixCipher@cpe-24-28-7-59.austin.res.rr.com] has joined ##hplusroadmap | 01:20 | |
| -!- drewbot [~cinch@ec2-54-83-103-185.compute-1.amazonaws.com] has quit [Ping timeout: 240 seconds] | 01:30 | |
| -!- drewbot [~cinch@ec2-54-87-28-12.compute-1.amazonaws.com] has joined ##hplusroadmap | 01:35 | |
| -!- jtimon [~quassel@103.red-80-26-235.adsl.dynamic.ccgg.telefonica.net] has joined ##hplusroadmap | 02:06 | |
| -!- vicarion [~net@2001:388:608c:6cb5:6ceb:f8d7:c2e3:39c5] has quit [Remote host closed the connection] | 02:06 | |
| -!- andares [~andares@2601:602:8f01:3150:488b:66:b9ac:9356] has joined ##hplusroadmap | 02:15 | |
| -!- andares [~andares@2601:602:8f01:3150:488b:66:b9ac:9356] has quit [Changing host] | 02:15 | |
| -!- andares [~andares@unaffiliated/jacco] has joined ##hplusroadmap | 02:15 | |
| -!- esmerelda [~andares@unaffiliated/jacco] has quit [Ping timeout: 250 seconds] | 02:19 | |
| -!- andares is now known as esmerelda | 02:40 | |
| -!- chris_99 [~chris_99@unaffiliated/chris-99/x-3062929] has joined ##hplusroadmap | 03:04 | |
| -!- FourFire [FourFire@cm-84.215.195.59.getinternet.no] has quit [Ping timeout: 245 seconds] | 03:20 | |
| -!- Gurkenglas [Gurkenglas@dslb-188-103-077-131.188.103.pools.vodafone-ip.de] has joined ##hplusroadmap | 03:29 | |
| -!- FourFire [FourFire@84.215.195.59] has joined ##hplusroadmap | 03:31 | |
| -!- esmerelda [~andares@unaffiliated/jacco] has quit [Ping timeout: 250 seconds] | 03:32 | |
| -!- sandeepkr [~sandeepkr@111.235.64.4] has quit [Ping timeout: 255 seconds] | 03:37 | |
| -!- esmerelda [~andares@unaffiliated/jacco] has joined ##hplusroadmap | 03:55 | |
| -!- jtimon [~quassel@103.red-80-26-235.adsl.dynamic.ccgg.telefonica.net] has quit [Ping timeout: 256 seconds] | 03:58 | |
| -!- FourFire [FourFire@84.215.195.59] has quit [Quit: Leaving] | 04:05 | |
| -!- QuadIgni [~fourfire@81.4.122.176] has joined ##hplusroadmap | 04:05 | |
| -!- rhaps0dy [~rhaps0dy@81.4.122.176] has joined ##hplusroadmap | 04:06 | |
| -!- QuadIgni is now known as FourFire | 04:10 | |
| -!- chris_99 [~chris_99@unaffiliated/chris-99/x-3062929] has quit [Ping timeout: 250 seconds] | 04:28 | |
| -!- jaboja [~jaboja@acvs86.neoplus.adsl.tpnet.pl] has joined ##hplusroadmap | 05:54 | |
| -!- jaboja [~jaboja@acvs86.neoplus.adsl.tpnet.pl] has quit [Ping timeout: 256 seconds] | 06:05 | |
| -!- Gurkenglas [Gurkenglas@dslb-188-103-077-131.188.103.pools.vodafone-ip.de] has quit [Ping timeout: 272 seconds] | 06:11 | |
| -!- sandeepkr [~sandeepkr@111.235.64.4] has joined ##hplusroadmap | 06:44 | |
| -!- vicarion [~net@2001:388:608c:6cb5:3d28:ec6b:ab06:70a4] has joined ##hplusroadmap | 06:50 | |
| -!- drewbot [~cinch@ec2-54-87-28-12.compute-1.amazonaws.com] has quit [Ping timeout: 260 seconds] | 07:01 | |
| kanzure | hmph | 07:02 |
|---|---|---|
| -!- drewbot [~cinch@ec2-54-226-123-45.compute-1.amazonaws.com] has joined ##hplusroadmap | 07:09 | |
| maaku | https://www.reddit.com/r/MachineLearning/comments/404r9m/ama_the_openai_research_team/ | 07:26 |
| maaku | .title | 07:26 |
| yoleaux | AMA: the OpenAI Research Team : MachineLearning | 07:26 |
| maaku | Apparently OpenAI will not be open source. Color me disappointed. | 07:28 |
| kanzure | so much for "everyone gets an ai" | 07:36 |
| kanzure | ai people are really really bad at "this time it's different" | 07:37 |
| -!- CheckDavid [uid14990@gateway/web/irccloud.com/x-qasdupqoyqefldsc] has joined ##hplusroadmap | 08:00 | |
| -!- erasmus [~erasmus@unaffiliated/erasmus] has joined ##hplusroadmap | 08:30 | |
| -!- abetusk [~abe@2601:184:4002:2290:69e6:24ea:434d:5782] has quit [Ping timeout: 260 seconds] | 08:31 | |
| -!- abetusk [~abe@2601:184:4002:2290:6c0c:ac9d:f404:4a2d] has joined ##hplusroadmap | 08:41 | |
| -!- abetusk is now known as Guest33754 | 08:42 | |
| -!- jtimon [~quassel@103.red-80-26-235.adsl.dynamic.ccgg.telefonica.net] has joined ##hplusroadmap | 08:44 | |
| -!- eudoxia [~eudoxia@r186-55-162-161.dialup.adsl.anteldata.net.uy] has joined ##hplusroadmap | 08:50 | |
| AdrianG | so how is it Open | 08:55 |
| -!- eudoxia [~eudoxia@r186-55-162-161.dialup.adsl.anteldata.net.uy] has quit [Quit: Leaving] | 08:56 | |
| -!- dustinm [~dustinm@105.ip-167-114-152.net] has quit [Ping timeout: 244 seconds] | 09:23 | |
| -!- Houshalter [~Houshalte@oh-71-50-56-224.dhcp.embarqhsd.net] has joined ##hplusroadmap | 09:25 | |
| -!- dustinm [~dustinm@2607:5300:100:200::160d] has joined ##hplusroadmap | 09:32 | |
| -!- atomical [~atomical@wl2142.cin.net] has joined ##hplusroadmap | 09:35 | |
| -!- erasmus [~erasmus@unaffiliated/erasmus] has quit [Quit: Namaste] | 09:44 | |
| -!- Pathseeker [~Pathseeke@176.4.135.231] has joined ##hplusroadmap | 09:57 | |
| -!- atomical [~atomical@wl2142.cin.net] has quit [Ping timeout: 260 seconds] | 09:59 | |
| archels | .title http://www.infoq.com/presentations/We-Really-Dont-Know-How-To-Compute | 10:00 |
| yoleaux | We Really Don't Know How To Compute! | 10:00 |
| archels | this man is saying so many things that are wrong that I don't even know where to begin | 10:00 |
| archels | and I'm scarcely 10 minutes in | 10:01 |
| JayDugger | People generally do "this time it's different" badly. | 10:01 |
| -!- Pathseeker [~Pathseeke@176.4.135.231] has quit [Ping timeout: 246 seconds] | 10:07 | |
| -!- atomical [~atomical@wl2142.cin.net] has joined ##hplusroadmap | 10:17 | |
| -!- chris_99 [~chris_99@unaffiliated/chris-99/x-3062929] has joined ##hplusroadmap | 10:51 | |
| fenn | maaku: i don't see any "official" responses on that reddit AMA; where you get "OpenAI will not be open source" from? | 10:58 |
| -!- Gurkenglas [Gurkenglas@dslb-188-103-077-131.188.103.pools.vodafone-ip.de] has joined ##hplusroadmap | 11:05 | |
| -!- CheckDavid [uid14990@gateway/web/irccloud.com/x-qasdupqoyqefldsc] has quit [Quit: Connection closed for inactivity] | 11:19 | |
| -!- |nodux [uid125132@gateway/web/irccloud.com/x-gsevlfbklrvuqymv] has joined ##hplusroadmap | 11:33 | |
| -!- jaboja [~jaboja@acvs86.neoplus.adsl.tpnet.pl] has joined ##hplusroadmap | 11:49 | |
| -!- chris_99 [~chris_99@unaffiliated/chris-99/x-3062929] has quit [Quit: Leaving] | 12:04 | |
| -!- Houshalter [~Houshalte@oh-71-50-56-224.dhcp.embarqhsd.net] has quit [Ping timeout: 256 seconds] | 12:09 | |
| -!- PatrickRobotham [uid18270@gateway/web/irccloud.com/x-iqkkiucwxvoukxxv] has quit [Quit: Connection closed for inactivity] | 12:10 | |
| -!- augur [~augur@c-73-46-94-9.hsd1.fl.comcast.net] has quit [Ping timeout: 240 seconds] | 12:22 | |
| -!- jaboja [~jaboja@acvs86.neoplus.adsl.tpnet.pl] has quit [Ping timeout: 250 seconds] | 12:24 | |
| -!- augur [~augur@c-73-46-94-9.hsd1.fl.comcast.net] has joined ##hplusroadmap | 12:47 | |
| xentrac | fenn: “I’d discussed with a staff member that the project would not in fact be open source” | 12:54 |
| -!- Houshalter [~Houshalte@oh-71-50-56-224.dhcp.embarqhsd.net] has joined ##hplusroadmap | 13:15 | |
| -!- chris_99 [~chris_99@unaffiliated/chris-99/x-3062929] has joined ##hplusroadmap | 13:17 | |
| -!- Houshalter [~Houshalte@oh-71-50-56-224.dhcp.embarqhsd.net] has quit [Ping timeout: 250 seconds] | 13:29 | |
| -!- andares [~andares@unaffiliated/jacco] has joined ##hplusroadmap | 13:39 | |
| -!- |nodux [uid125132@gateway/web/irccloud.com/x-gsevlfbklrvuqymv] has quit [Quit: Connection closed for inactivity] | 13:41 | |
| -!- esmerelda [~andares@unaffiliated/jacco] has quit [Ping timeout: 240 seconds] | 13:41 | |
| -!- |nodux [uid125132@gateway/web/irccloud.com/x-mcvszqfydlimozmk] has joined ##hplusroadmap | 13:50 | |
| -!- chris_99 [~chris_99@unaffiliated/chris-99/x-3062929] has quit [Quit: Leaving] | 14:00 | |
| -!- erasmus [~erasmus@unaffiliated/erasmus] has joined ##hplusroadmap | 14:12 | |
| -!- NixCipher [~NixCipher@cpe-24-28-7-59.austin.res.rr.com] has quit [Ping timeout: 264 seconds] | 14:16 | |
| -!- NixCipher [~NixCipher@76.164.234.138] has joined ##hplusroadmap | 14:32 | |
| -!- esmerelda [~andares@unaffiliated/jacco] has joined ##hplusroadmap | 14:41 | |
| -!- pompolic [~A@unaffiliated/pompolic] has quit [Ping timeout: 246 seconds] | 14:41 | |
| -!- pompolic [~A@unaffiliated/pompolic] has joined ##hplusroadmap | 14:43 | |
| -!- andares [~andares@unaffiliated/jacco] has quit [Ping timeout: 260 seconds] | 14:45 | |
| -!- Gurkenglas [Gurkenglas@dslb-188-103-077-131.188.103.pools.vodafone-ip.de] has quit [Ping timeout: 276 seconds] | 14:49 | |
| -!- jtimon [~quassel@103.red-80-26-235.adsl.dynamic.ccgg.telefonica.net] has quit [Ping timeout: 246 seconds] | 14:51 | |
| -!- Houshalter [~Houshalte@oh-71-50-56-224.dhcp.embarqhsd.net] has joined ##hplusroadmap | 14:52 | |
| -!- atomical [~atomical@wl2142.cin.net] has quit [Quit: My Mac has gone to sleep. ZZZzzz…] | 14:55 | |
| -!- atomical [~atomical@wl2142.cin.net] has joined ##hplusroadmap | 14:56 | |
| -!- sandeepkr [~sandeepkr@111.235.64.4] has quit [Ping timeout: 246 seconds] | 14:58 | |
| -!- atomical [~atomical@wl2142.cin.net] has quit [Quit: Textual IRC Client: www.textualapp.com] | 15:15 | |
| -!- NixCiph3r [~NixCipher@76.164.234.50] has joined ##hplusroadmap | 15:17 | |
| -!- NixCipher [~NixCipher@76.164.234.138] has quit [Ping timeout: 250 seconds] | 15:20 | |
| -!- vicarion [~net@2001:388:608c:6cb5:3d28:ec6b:ab06:70a4] has quit [Remote host closed the connection] | 15:27 | |
| -!- NixC1ph3r [~NixCipher@76.164.234.50] has joined ##hplusroadmap | 15:38 | |
| -!- NixCiph3r [~NixCipher@76.164.234.50] has quit [Ping timeout: 272 seconds] | 15:41 | |
| -!- hehelleshin is now known as helleshin | 15:42 | |
| -!- NixCiph3r [~NixCipher@108.62.56.66] has joined ##hplusroadmap | 15:54 | |
| -!- NixC1ph3r [~NixCipher@76.164.234.50] has quit [Ping timeout: 255 seconds] | 15:58 | |
| -!- NixCiph3r [~NixCipher@108.62.56.66] has quit [Ping timeout: 245 seconds] | 16:01 | |
| -!- |nodux [uid125132@gateway/web/irccloud.com/x-mcvszqfydlimozmk] has quit [Quit: Connection closed for inactivity] | 16:11 | |
| -!- NixCiph3r [~NixCipher@108.62.56.66] has joined ##hplusroadmap | 16:14 | |
| -!- Madplatypus [uid19957@gateway/web/irccloud.com/x-hjclelzsqrturosv] has quit [Quit: Connection closed for inactivity] | 16:27 | |
| -!- NixCiph3r [~NixCipher@108.62.56.66] has quit [Ping timeout: 255 seconds] | 16:41 | |
| maaku | fenn: what xentrac quoted | 16:43 |
| -!- vicarion [~net@2001:388:608c:6cb5:dd40:96e0:2017:da6f] has joined ##hplusroadmap | 16:49 | |
| -!- PatrickRobotham [uid18270@gateway/web/irccloud.com/x-wkdiuzfqnwmkcstl] has joined ##hplusroadmap | 16:50 | |
| -!- NixCiph3r [~NixCipher@76.164.234.82] has joined ##hplusroadmap | 16:56 | |
| -!- NixCiph3r [~NixCipher@76.164.234.82] has quit [Client Quit] | 16:58 | |
| -!- NixCipher [~NixCipher@76.164.234.82] has joined ##hplusroadmap | 16:58 | |
| -!- chris_99 [~chris_99@unaffiliated/chris-99/x-3062929] has joined ##hplusroadmap | 17:18 | |
| -!- |nodux [uid125132@gateway/web/irccloud.com/x-whccbtwhbwdzupqi] has joined ##hplusroadmap | 17:47 | |
| -!- andares [~andares@2607:fb90:8095:c517:9f85:11ef:8bda:d8bb] has joined ##hplusroadmap | 17:47 | |
| -!- andares [~andares@2607:fb90:8095:c517:9f85:11ef:8bda:d8bb] has quit [Changing host] | 17:47 | |
| -!- andares [~andares@unaffiliated/jacco] has joined ##hplusroadmap | 17:47 | |
| -!- esmerelda [~andares@unaffiliated/jacco] has quit [Ping timeout: 260 seconds] | 17:51 | |
| -!- Diablo-D3 [~diablo@exelion.net] has quit [Ping timeout: 255 seconds] | 18:12 | |
| -!- Diablo-D3 [~diablo@exelion.net] has joined ##hplusroadmap | 18:13 | |
| -!- jdqx [~jdqx@2602:306:cc94:1950:ec85:1783:4e7f:724] has joined ##hplusroadmap | 18:15 | |
| -!- Diablo-D3 [~diablo@exelion.net] has quit [Ping timeout: 260 seconds] | 18:18 | |
| maaku | So turns out that OpenAI is just a machine learning research shop. Yawn. | 18:18 |
| -!- esmerelda [~andares@unaffiliated/jacco] has joined ##hplusroadmap | 18:18 | |
| -!- andares [~andares@unaffiliated/jacco] has quit [Ping timeout: 250 seconds] | 18:22 | |
| -!- Diablo-D3 [~diablo@45.59.64.5] has joined ##hplusroadmap | 18:24 | |
| -!- chris_99 [~chris_99@unaffiliated/chris-99/x-3062929] has quit [Quit: Leaving] | 18:31 | |
| -!- Diablo-D3 [~diablo@45.59.64.5] has quit [Ping timeout: 272 seconds] | 18:32 | |
| -!- Diablo-D3 [~diablo@exelion.net] has joined ##hplusroadmap | 18:34 | |
| kanzure | what were you expecting instead, though? | 18:35 |
| -!- c0rw1n is now known as c0rw|zZz | 18:37 | |
| maaku | I'll repeat some things I said in ##AGI | 18:45 |
| maaku | It seems like Musk, Altmann, et al are getting swindled. | 18:46 |
| AdrianG | why? | 18:46 |
| maaku | They claim this is about x-risk reduction -- I know, bear with me -- but the actual research agenda of OpenAI has very little impact on that | 18:46 |
| AdrianG | go on. | 18:47 |
| maaku | I think actually throwing this money at MIRI would have done more x-risk reduction than OpenAI, and that's saying something if you know my views on that | 18:47 |
| xentrac | HAHA | 18:48 |
| xentrac | oops | 18:48 |
| maaku | OpenAI doesn't seem to be setting itself up to pursue anything other than next generation ML neural networks, which at best are a small part of AGI (emphasis on G) and not the part which leads to "the control problem" | 18:48 |
| xentrac | I normally don't have a capslock! | 18:48 |
| AdrianG | musk/altmann invested in openai? | 18:49 |
| maaku | It feels a bit like the 'nanotech' switcharoo in the '90s, where Foresight et al convinced the world that molecular manufacturing was a transformative and scary technology | 18:49 |
| maaku | So major gov'ts poured billions of $ into basic research into 'nanotech' ... which got redefined along the way to mean something else and we got stain resistant pants | 18:50 |
| AdrianG | openAI is getting funding because sutskever | 18:50 |
| maaku | AdrianG: yes, a billion dollars | 18:51 |
| AdrianG | i would imagine plenty of people would be interesting funding his work | 18:51 |
| maaku | (in the form of ralatively illiquid stocks and such) | 18:51 |
| AdrianG | maaku: afaik that's been pledged, not disbursed | 18:51 |
| AdrianG | idk. if i had a billion to blow, i would like to keep tabs on what sutskever is doing. | 18:51 |
| AdrianG | i think thats the real goal. | 18:52 |
| xentrac | maaku: wasn't that in the 2000s? | 18:53 |
| kanzure | musk is getting more x-risk reduction through spacex. miri's only idea of reduction is "total world domination by a nanny ai". and moratoriums. | 18:53 |
| xentrac | I was walking behind a smoker on the sidewalk earlier today and I reflected on the fact that she was filling the air with nanoparticles | 18:54 |
| maaku | xentrac: you are correct. year 2000 : https://en.wikipedia.org/wiki/National_Nanotechnology_Initiative | 18:54 |
| kanzure | non-smokers are also exhaling nanoparticles | 18:54 |
| maaku | it's 2016 but still when I think "that was last decade" my brain fills in the 1990's | 18:55 |
| kanzure | jokes on you, me i'm timelocked to 2008 | 18:55 |
| xentrac | are they, kanzure? not very many | 18:56 |
| kanzure | microbes don't count as nanoparticles, do they.... hmm. | 18:56 |
| maaku | xentrac: I think the point is more how useless the definition of "nanotech" == "nanoscale" is | 18:56 |
| kanzure | what about viruses? | 18:57 |
| kanzure | viruses are often nanoparticles | 18:57 |
| xentrac | when people exhale viruses, they are usually embedded in drops of sputum | 18:58 |
| kanzure | bah | 18:58 |
| maaku | kanzure: as to what I expected, I somehow thought that Sutskever would be heading up the ML portion of a larger AGI project | 18:58 |
| maaku | I didn't realize this whole thing was actually just a sutskever's personal research budget | 18:59 |
| maaku | So, suddenly much less interesting | 18:59 |
| xentrac | maaku: what do you think about TensorFlow? | 18:59 |
| maaku | I was about to say that with TensorFlow we already are getting access to most of the coolest tech coming out of Google Brain | 19:00 |
| kanzure | what is the origin of the lack of neuroanatomy in these projects? | 19:01 |
| maaku | So now the people that gave us TensorFlow are getting Musk's money instead of Larry and Sergei's. Personally I'm not sure that's an improvement.. Musk prob. would have done cooler things | 19:01 |
| kanzure | apparently he wouldn't have done cooler things- he picked this instead. | 19:01 |
| -!- erasmus [~erasmus@unaffiliated/erasmus] has quit [Quit: Namaste] | 19:03 | |
| maaku | kanzure: I think many wouldn't see the relevance of neuroanatomy to their work | 19:04 |
| xentrac | I would say that with TensorFlow we are getting access to a little bit of the coolest tech coming out of Google Brain five years ago :] | 19:04 |
| kanzure | maaku: really? | 19:04 |
| kanzure | machine vision people seem to look at the visual cortex a bunch | 19:05 |
| maaku | kanzure: though maybe I am projecting because I do not -- I have no interest in replicating human brains (there is a cheaper way to do that), but rather building intelligent machines that would easily solve problems humans struggle with | 19:05 |
| kanzure | (although not all of them) | 19:05 |
| maaku | machines that complement humans rather than replace them | 19:05 |
| kanzure | maaku: did you see the stitched diagram i slapped together? http://diyhpl.us/~bryan/papers2/neuro/cognitiveconsilience/stitched-diagram.jpg | 19:05 |
| -!- esmerelda [~andares@unaffiliated/jacco] has quit [Ping timeout: 260 seconds] | 19:05 | |
| maaku | kanzure: they do, but they tend to ignore inconvenient facts | 19:05 |
| maaku | e.g. the vision subsystem injects input into middle layers, not the bottom layer like every deepl learning architecture | 19:06 |
| maaku | and "neural nets" actually don't have anything resembling the real properties of neurons... | 19:07 |
| xentrac | also I have the impression that this is not true: | 19:08 |
| kanzure | looking at neuroanatomy does not require replicating human brains | 19:08 |
| xentrac | maaku> So now the people that gave us TensorFlow are getting Musk's money instead of Larry and Sergey's | 19:08 |
| xentrac | I think most of the people who worked on TensorFlow are still at Google and still working on it | 19:08 |
| maaku | kanzure: sure, but why should you care about the structure of the human brain if you are trying to solve an entirely different problem? | 19:10 |
| kanzure | entirely different? | 19:10 |
| maaku | kanzure: Spekaing for myself, I'm interested in general intelligence that is able to solve problems humans are particularly bad at | 19:12 |
| maaku | architectural differences are a good thing in that respect | 19:12 |
| kanzure | so your plan to do this is to have humans solve a problem that solves human-unsolvable problems? | 19:12 |
| maaku | Yes | 19:13 |
| maaku | But unsolvable is the wrong word -- just hard. | 19:13 |
| maaku | E.g. something that takes 10,000 engineers working for 10 years might just take a reasonably sized datacenter 2 months. | 19:14 |
| maaku | In other words, productivity improvements. | 19:14 |
| maaku | Anyway. This is interesting watching from the outside because it resembles Blockstream .. a group of academics finds excessive funding to do basic research with a commercial component in a space a lot of people have emotional investments into. | 19:15 |
| maaku | It's somewhat enlightening to be on this side of it. | 19:15 |
| maaku | kanzure: I wonder what you would posit the defined goal of AGI would be? | 19:17 |
| kanzure | besides total world domination? | 19:17 |
| kanzure | human-level cognitive ability, including communication (i don't care if it doesn't pass the turing test) | 19:18 |
| maaku | Why communication? | 19:18 |
| kanzure | input/output is important | 19:19 |
| maaku | Ok so communication as a way to impart goals onto the device, and return results | 19:20 |
| kanzure | nah, just a way to help with debugging, a way for the software to acquire more information, a way to probe human knowledge | 19:20 |
| maaku | Idk, my vision has always been some sort of formal language -- think Haskell, not English -- for user interaction and debugging | 19:21 |
| kanzure | well i mean, all software can be "debugged" through traditional means | 19:22 |
| maaku | it would need to be able to comprehend written language to acquire knowledge, but e.g. what Watson does is just fine. It doesn't need human-replica language centers | 19:22 |
| kanzure | but inline/online feedback (through communication) is useful for other reasons | 19:22 |
| kanzure | watson doesn't do what you think it does | 19:22 |
| maaku | Watson-like, but that's all it needs to do I think. | 19:23 |
| maaku | All I would need it to do is provide heuristic guidance for evolutionary search. I think Watson is sufficient for that. | 19:25 |
| maaku | kanzure: We digress. I don't think building AGI to do things people are already particularly good at is very interesting, although it seems to interest many other people. | 19:27 |
| maaku | I see it like this: imagine we had to build a nuclear reactor in 7-dimensional space. We *could* do that just by working out and trusting the math, but we'd not be very good at it | 19:28 |
| maaku | Progress would be slow to say the least. AGI shines here because we can build an artificial nuclear engineer with intuition for 7-dimensional spaces | 19:28 |
| maaku | Back to reality, I see parallels to both biology and molecular nanotechnology. For different reasons I think in both cases our efforts are hampered by deficiencies in the human mind | 19:30 |
| maaku | In the case of biology is is simple the insane number of processes going on and everything interconnecting with everything else ... computer mediated research is the only way to progress, and AGI lets you take human biologists out of the loop so you can really progress at speed | 19:31 |
| xentrac | maaku: user interaction in Lojban? | 19:35 |
| maaku | xentrac: actually sortof yes. Not Lojban per se, but something like it | 19:36 |
| xentrac | why not Lojban per se? | 19:36 |
| maaku | it's a poor fit. it's basically first-order logic and doesn't natively support the sorts of constructions that would be necessary to efficiently convey information | 19:37 |
| xentrac | oh, interesting | 19:37 |
| maaku | but a logical language with constructs taken from probabalistic graphical models would fit the bill | 19:38 |
| xentrac | have you thought about outlining such a thing? I have no idea what it would look like | 19:38 |
| maaku | it's been on my list of things to do for years | 19:39 |
| maaku | definately not the long pole in the tent at the moment however | 19:39 |
| maaku | I've tried to interest others in taking a stab at it, without success | 19:39 |
| maaku | Attacking it from the other direction, this is also what you would get if you took probabalistic programming languages and added context and ambiguity | 19:43 |
| xentrac | I mean it sounds like a really interesting idea but I don't understand your idea well enough to take a stab at it | 19:44 |
| -!- ArturShaik [~ArturShai@212.241.15.137] has quit [Ping timeout: 265 seconds] | 19:51 | |
| -!- |nodux is now known as |node | 19:54 | |
| -!- erasmus [~erasmus@unaffiliated/erasmus] has joined ##hplusroadmap | 20:04 | |
| -!- ArturShaik [~ArturShai@212.97.21.162] has joined ##hplusroadmap | 20:05 | |
| -!- CheckDavid [uid14990@gateway/web/irccloud.com/x-uappinxgwfuyahtj] has joined ##hplusroadmap | 20:14 | |
| -!- vicarion [~net@2001:388:608c:6cb5:dd40:96e0:2017:da6f] has quit [Remote host closed the connection] | 20:27 | |
| -!- esmerelda [~andares@2601:602:8f01:3150:5472:e4ce:bb69:f483] has joined ##hplusroadmap | 20:33 | |
| -!- esmerelda [~andares@2601:602:8f01:3150:5472:e4ce:bb69:f483] has quit [Changing host] | 20:33 | |
| -!- esmerelda [~andares@unaffiliated/jacco] has joined ##hplusroadmap | 20:33 | |
| -!- Houshalter [~Houshalte@oh-71-50-56-224.dhcp.embarqhsd.net] has quit [Ping timeout: 276 seconds] | 20:49 | |
| -!- Houshalter [~Houshalte@oh-71-50-56-224.dhcp.embarqhsd.net] has joined ##hplusroadmap | 20:53 | |
| -!- andares [~andares@c-73-225-80-196.hsd1.wa.comcast.net] has joined ##hplusroadmap | 20:58 | |
| -!- andares [~andares@c-73-225-80-196.hsd1.wa.comcast.net] has quit [Changing host] | 20:58 | |
| -!- andares [~andares@unaffiliated/jacco] has joined ##hplusroadmap | 20:58 | |
| -!- esmerelda [~andares@unaffiliated/jacco] has quit [Read error: Connection reset by peer] | 20:58 | |
| -!- andares [~andares@unaffiliated/jacco] has quit [Read error: Connection reset by peer] | 20:58 | |
| -!- esmerelda [~andares@2601:602:8f01:3150:5472:e4ce:bb69:f483] has joined ##hplusroadmap | 20:58 | |
| -!- esmerelda [~andares@2601:602:8f01:3150:5472:e4ce:bb69:f483] has quit [Changing host] | 20:58 | |
| -!- esmerelda [~andares@unaffiliated/jacco] has joined ##hplusroadmap | 20:58 | |
| -!- andares [~andares@2607:fb90:80b0:c2b5:4ae5:4274:423:a63c] has joined ##hplusroadmap | 21:00 | |
| -!- andares [~andares@2607:fb90:80b0:c2b5:4ae5:4274:423:a63c] has quit [Changing host] | 21:00 | |
| -!- andares [~andares@unaffiliated/jacco] has joined ##hplusroadmap | 21:00 | |
| -!- esmerelda [~andares@unaffiliated/jacco] has quit [Read error: Connection reset by peer] | 21:00 | |
| -!- esmerelda [~andares@unaffiliated/jacco] has joined ##hplusroadmap | 21:01 | |
| -!- andares [~andares@unaffiliated/jacco] has quit [Ping timeout: 250 seconds] | 21:05 | |
| -!- andares [~andares@2607:fb90:808c:db0f:583e:f78e:ce44:5378] has joined ##hplusroadmap | 21:05 | |
| -!- andares [~andares@2607:fb90:808c:db0f:583e:f78e:ce44:5378] has quit [Changing host] | 21:05 | |
| -!- andares [~andares@unaffiliated/jacco] has joined ##hplusroadmap | 21:05 | |
| -!- esmerelda [~andares@unaffiliated/jacco] has quit [Read error: Connection reset by peer] | 21:05 | |
| -!- esmerelda [~andares@unaffiliated/jacco] has joined ##hplusroadmap | 21:06 | |
| -!- andares [~andares@unaffiliated/jacco] has quit [Ping timeout: 250 seconds] | 21:10 | |
| -!- drewbot_ [~cinch@ec2-54-160-214-102.compute-1.amazonaws.com] has joined ##hplusroadmap | 21:11 | |
| -!- andares [~andares@2607:fb90:80b8:1f88:8f6b:ecf9:6712:85d] has joined ##hplusroadmap | 21:18 | |
| -!- andares [~andares@2607:fb90:80b8:1f88:8f6b:ecf9:6712:85d] has quit [Changing host] | 21:18 | |
| -!- andares [~andares@unaffiliated/jacco] has joined ##hplusroadmap | 21:18 | |
| -!- drewbot [~cinch@ec2-54-226-123-45.compute-1.amazonaws.com] has quit [Ping timeout: 276 seconds] | 21:18 | |
| -!- esmerelda [~andares@unaffiliated/jacco] has quit [Read error: Connection reset by peer] | 21:18 | |
| -!- esmerelda [~andares@unaffiliated/jacco] has joined ##hplusroadmap | 21:19 | |
| -!- andares [~andares@unaffiliated/jacco] has quit [Ping timeout: 240 seconds] | 21:22 | |
| -!- Madplatypus [uid19957@gateway/web/irccloud.com/x-twrtgikfbcghdpvx] has joined ##hplusroadmap | 21:24 | |
| -!- esmerelda [~andares@unaffiliated/jacco] has quit [Read error: Connection reset by peer] | 21:38 | |
| -!- esmerelda [~andares@unaffiliated/jacco] has joined ##hplusroadmap | 21:38 | |
| maaku | xentrac: The idea is quite simple. Most AGI projects involve some sort of knowledgebase, which is in latest incarnations a graph structure with probabalistic weights | 21:41 |
| maaku | This graph encodes knowledge, including relationships between concepts and meta-level data like how a concept was learned or what evidence there is for it | 21:42 |
| kanzure | i forgot how much i disagree with you about this topic | 21:43 |
| maaku | The language would be a direct linear serialization of this graph structure | 21:43 |
| maaku | kanzure: I find your own views equally puzzling | 21:43 |
| kanzure | you consider my views wrong, but i don't think you are puzzled by them | 21:44 |
| kanzure | i attribute some functionality to human brain matter that i am interested in either simulating or emulating through non-brain materials, and you simply are not... | 21:46 |
| maaku | no, I am confused. i do not consider it a wrong approach to work towards brain emulation. that is a necessary component of uploading | 21:46 |
| kanzure | it is also useful even in scenarios without uploading | 21:47 |
| maaku | yeah see that's what i'm confused about. to me it is like an aeronautical engineer in 1902 trying to make flying machines with flapping wings and feathers | 21:47 |
| kanzure | i hate bird airplane analogies. i absolutely fucking hate them. birds work, and so do airplanes. | 21:48 |
| maaku | yes, but airplanes are easier to design and reliably build than birds | 21:48 |
| maaku | s/build/engineer/ | 21:49 |
| kanzure | so? | 21:49 |
| kanzure | i also hate the analogy that computation is like fixed wing lift; it assumes the outcome. | 21:50 |
| maaku | I don't want to make human-identical intelligences. I've already done that twice. I want to solve prolem X in as highly leveraged a way as possible. | 21:50 |
| kanzure | we already know that computation is necessary for information processing capacity. it's a completely useless observation. | 21:50 |
| kanzure | so making "computation" fixed-wing is retarded; the brain is doing "computation" too, you know. | 21:51 |
| kanzure | there is no guarantee that brain simulation or brain emulation will produce "identical intelligence". | 21:51 |
| kanzure | so that does not seem like a concern..... | 21:51 |
| kanzure | also, you have done the physical manifestation biology brain thing; software as you know is much more highly scalable and deployable, testable and copyable. | 21:53 |
| kanzure | and you don't have to feel as bad about performing dangerous experiments on 'em | 21:53 |
| maaku | kanzure: Here's what I want: a nanofactory, a complete mapping of the human proteome and all related biological processes, and the ability to reliably make arbitrary changes thereof using said nanotechnology on the scale of days/weeks not years | 21:58 |
| maaku | oh, and I want it in the next 40 years or else it's a moot point anyway | 21:58 |
| kanzure | what do you consider a proteome map to be? | 21:59 |
| maaku | the complete expression of the human genome, including the general location, function, lifecycle, and failure modes of every protien | 22:00 |
| maaku | and by extension, how they come together to form cells, tissue, human beings | 22:01 |
| kanzure | so all protein-protein interactions yes/no? | 22:01 |
| kanzure | just curious | 22:01 |
| maaku | i suppose so yes | 22:02 |
| maaku | I don't actually give a damn about AI in any intrinsic way, unlike most people of the field. It just seems like the only viable pathway to achiving the above within the time constraints and without a trillion dollars. | 22:03 |
| maaku | "Solve the entire field of human biology" is a task I don't know how to even attempt in a single lifetime. | 22:04 |
| maaku | "Build an artificial biologist and then scale it" on the other hand seems plausible. | 22:05 |
| kanzure | your biological goals are surely not unlimited knowledge, but rather some specific outcomes, ya? | 22:06 |
| maaku | Yes. I mean unlimited capability to acquire knowledge, not all knowledge already aquired. | 22:07 |
| -!- andares [~andares@2607:fb90:8091:c93:c449:ff45:e961:6429] has joined ##hplusroadmap | 22:09 | |
| -!- andares [~andares@2607:fb90:8091:c93:c449:ff45:e961:6429] has quit [Changing host] | 22:09 | |
| -!- andares [~andares@unaffiliated/jacco] has joined ##hplusroadmap | 22:09 | |
| maaku | Human beings aren't actually good at doing science. It took a heck of a long time to figure it out as a species, and even individually it takes many years of grad school or practical experience to be effective. | 22:11 |
| -!- esmerelda [~andares@unaffiliated/jacco] has quit [Ping timeout: 245 seconds] | 22:12 | |
| kanzure | most of that effectiveness delay is due to incredibly high inefficiencies in the existing education system. it's not like the system was designed for optimization..... see john gatto stuff. | 22:12 |
| maaku | That is true, but our brains are also pretty shitty reasoning engines, much better suited to forming superstitious beliefs or replaying cached wrong answers | 22:13 |
| maaku | People like you and me and the other fine people of this channel and the 0.01% of people that become truly impactful scientists or engineers are the exceptions (numbers pulled out of hat) | 22:14 |
| maaku | And not because we were born that way but because we beat the odds and forced ourselves to learn proper habits re: rigor and such | 22:15 |
| -!- sandeepkr [~sandeepkr@111.235.64.4] has joined ##hplusroadmap | 22:16 | |
| maaku | So, my thesis: it's probably a better idea to look at how an ideal artificial scientist would reason, and implement that, than try to smash together a bunch of biologically inspired neural networks to make another error-prone superstitious reasoner | 22:16 |
| nmz787 | maaku: idk about "not being good at science" if I was some scientist's apprentice starting when I was 8 or 10, I would have been at the prime age to start working on ideas | 22:18 |
| nmz787 | ideas/problems | 22:18 |
| -!- CheckDavid [uid14990@gateway/web/irccloud.com/x-uappinxgwfuyahtj] has quit [Quit: Connection closed for inactivity] | 22:19 | |
| maaku | nmz787: Idk how old you are but I assume you're an adult. So you just said it would take decades of apprenticeship to become an effective scientist. | 22:20 |
| maaku | If it was actually something we were truly good at, it'd be like breathing. | 22:21 |
| maaku | Or hey, something reasonably complex: walking. | 22:21 |
| maaku | It's an irony of psychology: we only think about and are conscious of the things we are not good at. We think hard about the stuff we are terrible at. | 22:22 |
| kanzure | superstitious brain matter seems to do a lot of really interesting things that we don't know how to do on computers yet | 22:25 |
| -!- jdqx [~jdqx@2602:306:cc94:1950:ec85:1783:4e7f:724] has quit [Read error: Connection reset by peer] | 22:49 | |
| -!- jdqx [~jdqx@2602:306:cc94:1950:ec85:1783:4e7f:724] has joined ##hplusroadmap | 22:49 | |
| -!- esmerelda [~andares@2601:602:8f01:3150:5472:e4ce:bb69:f483] has joined ##hplusroadmap | 22:49 | |
| -!- esmerelda [~andares@2601:602:8f01:3150:5472:e4ce:bb69:f483] has quit [Changing host] | 22:49 | |
| -!- esmerelda [~andares@unaffiliated/jacco] has joined ##hplusroadmap | 22:49 | |
| -!- andares [~andares@unaffiliated/jacco] has quit [Ping timeout: 250 seconds] | 22:53 | |
| nmz787 | maaku: no I was more saying I would have been an effective apprentice-scientist right away at that age. | 22:56 |
| nmz787 | maaku: we didn't really need to think about walking before we were ready/able to start replicating it (i.e. when materials light and strong enough and motors/pneumatics/hydraulics that had enough power density, etc) | 23:01 |
| nmz787 | we were thinking a lot about weapons to kill wolves | 23:01 |
| nmz787 | or for food | 23:01 |
| nmz787 | or maybe some medical doctors over the years thought about it | 23:02 |
| -!- erasmus [~erasmus@unaffiliated/erasmus] has quit [Quit: Namaste] | 23:03 | |
| nmz787 | regarding surgery and birth anomalies | 23:03 |
| nmz787 | or disease | 23:03 |
| -!- c0rw|zZz [~c0rw1n@91.176.76.47] has quit [Ping timeout: 260 seconds] | 23:06 | |
| -!- |node [uid125132@gateway/web/irccloud.com/x-whccbtwhbwdzupqi] has quit [Quit: Connection closed for inactivity] | 23:31 | |
| -!- sandeepkr [~sandeepkr@111.235.64.4] has quit [Ping timeout: 256 seconds] | 23:46 | |
| -!- jaboja [~jaboja@erv84.neoplus.adsl.tpnet.pl] has joined ##hplusroadmap | 23:53 | |
| --- Log closed Sun Jan 10 00:00:00 2016 | ||
Generated by irclog2html.py 2.15.0.dev0 by Marius Gedminas - find it at mg.pov.lt!
