--- Log opened Fri Apr 22 00:00:02 2022
03:13 -!- rndhouse [rndhouse@gateway/vpn/protonvpn/rndhouse] has joined #hplusroadmap
04:21 -!- darsie [~darsie@84-113-55-200.cable.dynamic.surfer.at] has joined #hplusroadmap
05:06 -!- yashgaroth [~ffffffff@2607:fb90:a829:cb6f:599b:2830:1691:640] has joined #hplusroadmap
05:30 < lsneff> maaku: 100-200 kiloton
06:59 -!- mirage335 [~mirage335@64.79.52.86] has quit [Quit: Client closed]
06:59 -!- mirage335 [~mirage335@64.79.52.86] has joined #hplusroadmap
07:07 -!- mirage335 [~mirage335@64.79.52.86] has quit [Quit: Client closed]
07:07 -!- mirage335 [~mirage335@64.79.52.86] has joined #hplusroadmap
07:14 < docl> lsneff: awesome!
07:25 < docl> as I understand it there's not really a rigorously established lower bound, and in my personal opinion 100 kT is probably a significant overestimate (even if you assume it necessarily has elevators reaching to the ground). Paul Birch used 180 kT as an example of a small version, but he wasn't apparently trying to establish a lower bound
07:27 < docl> he contrasted that with a 180 million ton version to show the desirability of bootstrapping. unfortunately people skimming his paper got the impression that the 31 trillion dollar price tag he calculated for that straw example was the actual projected cost. wikipedia carried this error for several years before I corrected it
07:57 < docl> this is also wrong: https://space.nss.org/settlement/nasa/Nowicki/SPBI127.HTM
07:57 < docl> .title
07:57 < saxo> Orbital Ring
11:13 < kanzure> .tw https://twitter.com/erika_alden_d/status/1517402802045132800
11:13 < saxo> Alexis made a de novo protein that MOVES 🤯 I maintain this is the best project the Baker lab has ever done. https://www.science.org/doi/10.1126/science.abm1183#.YmJWpa4qnAA.twitter (@erika_alden_d)
11:13 < kanzure> https://www.science.org/doi/10.1126/science.abm1183
11:13 < kanzure> .tw https://twitter.com/bayeslord/status/1517237945010327553
11:13 < saxo> apologies to oxford, the nation of australia, and the entire singer family https://pbs.twimg.com/media/FQ5P3nPVkAEaE9e.jpg (@bayeslord)
11:14 < kanzure> .tw https://twitter.com/wootsecurity/status/1517495940206235648
11:14 < saxo> Up for an rootkit implementation on ARM's TrustZone? Read this! "Abusing Trust: Mobile Kernel Subversion via TrustZone Rootkits" by Daniel Marth et al. @ESSE_research @reed_solomon https://pbs.twimg.com/media/FQSkc4-VQActgIR.jpg (@wootsecurity)
11:14 < kanzure> .tw https://twitter.com/neuroecology/status/1516565594283384837
11:14 < saxo> You can plant an undetectable backdoor in any deep learning model / https://arxiv.org/abs/2204.06974 https://pbs.twimg.com/media/FQvsRuqX0Asg_k9.jpg (@neuroecology)
11:14 < muurkha> read a jpeg?
11:15 < superkuh> https://arxiv.org/abs/2204.06974
11:15 < muurkha> .t
11:15 < superkuh> Oh, the link was already there.
11:16 < saxo> [2204.06974] Planting Undetectable Backdoors in Machine Learning Models
11:16 < muurkha> I meant the rootkit
11:19 < superkuh> Ah.
12:04 -!- spaceangel [~spaceange@ip-78-102-216-202.net.upcbroadband.cz] has joined #hplusroadmap
14:28 -!- spaceangel [~spaceange@ip-78-102-216-202.net.upcbroadband.cz] has quit [Remote host closed the connection]
14:36 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=db656d49 Bryan Bishop: transcript: another CTV workshop >> http://diyhpl.us/diyhpluswiki/transcripts/ctv-bip-review-workshop-2022/
15:07 -!- rndhouse [rndhouse@gateway/vpn/protonvpn/rndhouse] has quit [Quit: Leaving]
16:03 < fenn> why is "the nation of australia" in that list?
16:09 < fenn> jesus christ, so much for "content filtering" (thread) https://nitter.net/bayeslord/status/1511191083341996032#m
16:10 < fenn> "Only be eradicating all life can we be sure that no one will suffer or die needlessly."
16:10 < fenn> by*
16:12 < fenn> sorta can't believe what i'm reading here
16:21 < muurkha> that's a well-known consequence of negative utilitarianism, fenn
16:21 < muurkha> the benevolent world-exploder
16:21 < fenn> yeah but it's just casually thrown in as the justification for some evil villain character
16:22 < muurkha> seems reasonable, why does that surprise you?
16:26 < fenn> uh, well, i guess i'm used to generated text being sorta "not all there" and a random mishmash of related things, whereas this is a coherent and justified system of thoughts
16:27 < fenn> i doubt you will find many human mass murderers with such a well justified motive
16:27 < muurkha> oh, is this an output from GPT-3 or something?  all I see is "hm, off to a grandiose start"
16:27 < fenn> the conversation text is screenshots
16:27 < muurkha> yeah I see that there are JPEGs
16:27 < fenn> i'm sorry that you cannot read
16:28 < muurkha> it'll be okay
16:28 < muurkha> anyway it's quite likely that the benevolent world exploder is in the training set if you're looking at a text predictive model
16:31 < fenn> now i am worrying about whether to type this into google :(
16:31 < fenn> i'm not seeing any exact substring matches, but, well, google sucks now
16:33 < muurkha> generative text models are pretty good at paraphrasing things since GPT-2 tho
16:35 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has left #hplusroadmap []
16:46 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has joined #hplusroadmap
18:49 < fenn> the generated paintings of robots with flowers popping out of their heads, which i posted yesterday as a video, now here as simply 1000 jpg files (the 6 extra are from the cover and inset pages i guess)
18:50 < fenn> http://fennetic.net/1000_robots/
19:08 < superkuh> Neat.
19:11 < fenn> maybe some day i'll do like a clickable gallery with sections for each prompt
19:11 < fenn> but for now it exists
19:14 -!- yashgaroth [~ffffffff@2607:fb90:a829:cb6f:599b:2830:1691:640] has quit [Quit: Leaving]
19:36 -!- xaete [~user12345@2607:9880:2458:58:4d8e:58f1:cf2b:45ae] has joined #hplusroadmap
20:57 < lsneff> Pattern matching genocide
21:15 -!- darsie [~darsie@84-113-55-200.cable.dynamic.surfer.at] has quit [Ping timeout: 256 seconds]
23:31 < maaku> .title https://phys.org/news/2022-04-axle-rotor-nanomachine.html
23:31 < saxo> Researchers take a step toward creating an axle-rotor nanomachine
--- Log closed Sat Apr 23 00:00:03 2022