--- Log opened Sun Jul 30 00:00:35 2017
00:01 -!- aeiousomething [~aeiousome@124.123.15.66] has joined ##hplusroadmap
00:42 -!- yashgaroth [~yashgarot@2606:6000:cd4d:3300:f5e0:f867:a11d:8d52] has quit [Quit: Leaving]
00:53 < nmz787> hmm, a new video card could test this out https://github.com/mozilla/DeepSpeech
00:54 < nmz787> .title
00:54 < yoleaux> GitHub - mozilla/DeepSpeech: A TensorFlow implementation of Baidu's DeepSpeech architecture
01:19 < archels_> just ordered some bromantane for trialling
02:24 < maaku> kanzure: heat the atmosphere to 100c?
02:29 < maaku> Why does George Church, Daniel Dennett, and James Randi all look alike? I want a movie where they're all played by the same actor.
02:33 < fltrz> because of time travel, they're the same person
02:34 < fltrz> steals peoples ideas and goes back in time
02:42 -!- Malvolio [~Malvolio@unaffiliated/malvolio] has quit [Ping timeout: 240 seconds]
03:01 -!- aeiousomething [~aeiousome@124.123.15.66] has quit [Ping timeout: 240 seconds]
03:40 < kanzure> accusing them of plagiarism is serious business, you might be next on their list now. you sure you want that?
03:41 < kanzure> mozilla/deepspeech is very intereting. i had written an implementation of deepspeech using keras and tensorflow, but i'll gladly take mozilla's reimplementation attempt over mine.
03:42 < kanzure> here is someone attempting to use it with a language model https://github.com/mozilla/DeepSpeech/issues/720
03:42 < kanzure> cc maaku
04:01 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has joined ##hplusroadmap
04:05 -!- emeraldgreen [~user@92.222.68.248] has joined ##hplusroadmap
04:05 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has quit [Ping timeout: 260 seconds]
04:06 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has joined ##hplusroadmap
04:32 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has quit [Ping timeout: 276 seconds]
05:05 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has joined ##hplusroadmap
05:13 -!- preview [~quassel@2407:7000:842d:4000::3] has quit [Read error: Connection reset by peer]
05:15 -!- Guest75407 [~talinck@cpe-174-97-113-184.cinci.res.rr.com] has quit [Ping timeout: 260 seconds]
05:16 -!- helleshin [~talinck@cpe-174-97-113-184.cinci.res.rr.com] has joined ##hplusroadmap
05:48 -!- aeiousomething [~aeiousome@124.123.15.66] has joined ##hplusroadmap
05:57 < fltrz> kanzure he only steals from the best, so that rules me out ;)
05:59 -!- aeiousomething [~aeiousome@124.123.15.66] has quit [Ping timeout: 240 seconds]
06:00 -!- aeiousomething [~aeiousome@124.123.15.66] has joined ##hplusroadmap
06:17 -!- Gurkenglas_ [~Gurkengla@dslb-178-005-166-215.178.005.pools.vodafone-ip.de] has joined ##hplusroadmap
06:35 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has quit [Remote host closed the connection]
06:54 -!- mf1008 [~mf1008@unaffiliated/mf1008] has quit [Quit: Cave quid dicis, quando, et cui]
06:55 -!- mf1008 [~mf1008@unaffiliated/mf1008] has joined ##hplusroadmap
06:55 -!- emeraldgreen [~user@92.222.68.248] has quit [Quit: Leaving.]
06:58 -!- aeiousomething [~aeiousome@124.123.15.66] has quit [Ping timeout: 240 seconds]
07:33 -!- aeiousomething [~aeiousome@124.123.15.66] has joined ##hplusroadmap
07:40 -!- darsie [~darsie@84-113-55-42.cable.dynamic.surfer.at] has quit [Ping timeout: 260 seconds]
08:00 -!- aeiousomething [~aeiousome@124.123.15.66] has quit [Ping timeout: 255 seconds]
09:27 -!- Malvolio [~Malvolio@unaffiliated/malvolio] has joined ##hplusroadmap
09:33 -!- nos [~nos@37-136-201-169.rev.dnainternet.fi] has joined ##hplusroadmap
09:33 < nos> Here I am! \o/
09:41 -!- Douhet [~Douhet@unaffiliated/douhet] has quit [Ping timeout: 268 seconds]
09:42 -!- Douhet [~Douhet@unaffiliated/douhet] has joined ##hplusroadmap
09:48 -!- Douhet [~Douhet@unaffiliated/douhet] has quit [Ping timeout: 246 seconds]
09:48 -!- Douhet [~Douhet@unaffiliated/douhet] has joined ##hplusroadmap
09:52 < archels_> welcome
09:54 -!- aeiousomething [~aeiousome@124.123.15.66] has joined ##hplusroadmap
09:55 -!- JayDugger [~jwdugger@47.185.237.246] has quit [Quit: Leaving.]
10:07 < nos> Found out about this channel from a comment on HaD.
10:07 < kanzure> greetings.
10:11 -!- TinKode [~TinKode@unaffiliated/tinkode] has joined ##hplusroadmap
10:17 < nos> so... have you guys heard about quantum radar?
10:18 < kanzure> .wik quantum radar
10:18 < yoleaux> "Quantum radar is a remote-sensing method based on quantum entanglement. The most convincing model has been proposed by an international team of researchers." — https://en.wikipedia.org/wiki/Quantum_radar
10:22 < nos> You could possibly get the essential ingredient on eBay, in the form of YIG sphere filters. 50 to 150 USD a pop.
10:23 < nos> And a friend tells me this could be used for protein spectrometry.
10:23 < nos> ~spectroscopy
10:25 < kanzure> instead of a spectrophotometer?
10:27 < nos> I guess so. My friend is into neuroscience. I'm into non-bio tech.
10:33 < kanzure> biology is the oldest technology
10:38 < nos> One day I will replace my bio-friends with robot friends.
10:39 -!- marciogm [~xxxIdent@unaffiliated/marciogm] has quit [Ping timeout: 246 seconds]
11:21 -!- Douhet [~Douhet@unaffiliated/douhet] has quit [Ping timeout: 246 seconds]
11:23 -!- augur [~augur@noisebridge130.static.monkeybrains.net] has joined ##hplusroadmap
11:26 -!- Douhet [~Douhet@unaffiliated/douhet] has joined ##hplusroadmap
11:30 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has joined ##hplusroadmap
11:33 -!- deep-book-gk_ [~1wm_su@119.81.29.178] has joined ##hplusroadmap
11:34 -!- deep-book-gk_ [~1wm_su@119.81.29.178] has left ##hplusroadmap []
11:35 -!- poppingtonic1 [~brian@51.15.130.225] has joined ##hplusroadmap
11:36 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has quit [Remote host closed the connection]
11:49 -!- Gurkenglas__ [~Gurkengla@dslb-178-005-166-215.178.005.pools.vodafone-ip.de] has joined ##hplusroadmap
11:50 -!- Douhet [~Douhet@unaffiliated/douhet] has quit [Ping timeout: 246 seconds]
11:50 -!- Douhet [~Douhet@unaffiliated/douhet] has joined ##hplusroadmap
11:52 -!- Gurkenglas_ [~Gurkengla@dslb-178-005-166-215.178.005.pools.vodafone-ip.de] has quit [Ping timeout: 255 seconds]
11:59 -!- scott__ [~scott@host-128-227-196-212.xlate.ufl.edu] has left ##hplusroadmap ["Leaving"]
12:01 -!- poppingtonic1 [~brian@51.15.130.225] has quit [Read error: Connection reset by peer]
12:15 -!- CheckDavid [uid14990@gateway/web/irccloud.com/x-zkcevvguekhuyqtc] has joined ##hplusroadmap
12:17 -!- hehelleshin [~talinck@cpe-174-97-113-184.cinci.res.rr.com] has joined ##hplusroadmap
12:19 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has joined ##hplusroadmap
12:21 -!- helleshin [~talinck@cpe-174-97-113-184.cinci.res.rr.com] has quit [Ping timeout: 260 seconds]
12:38 -!- Douhet [~Douhet@unaffiliated/douhet] has quit [Ping timeout: 268 seconds]
12:39 -!- Douhet [~Douhet@unaffiliated/douhet] has joined ##hplusroadmap
12:40 -!- aeiousomething [~aeiousome@124.123.15.66] has quit [Ping timeout: 246 seconds]
12:50 < kanzure> https://github.com/openmicroscopy
13:05 < pasky> nos: thumbs up! ;)
13:16 < kanzure> reason@fightaging.org commenting about juvenescence company: https://news.ycombinator.com/item?id=14887534
13:21 < nmz787> kanzure: what CPU, GPU and RAM are/were you using for your DeepSpeech tests?
13:22 < kanzure> whatever i picked on amazon aws.
13:23 < kanzure> 4 GB gpu of some kind.
13:25 < nmz787> can you find the model?
13:28 < kanzure> g2.4xlarge i think.
13:28 < kanzure> click "advanced computing" to see aws current generation of instance types https://aws.amazon.com/ec2/instance-types/
13:28 < kanzure> er, "accelerated computing"
13:29 < kanzure> iirc it was nvidia k80 gpu
13:30 < kanzure> get something pretty for yourself but i think google cloud tensorflow thingy is competitively priced
13:39 -!- yashgaroth [~yashgarot@2606:6000:cd4d:3300:f5e0:f867:a11d:8d52] has joined ##hplusroadmap
13:50 -!- JayDugger [~jwdugger@47.185.237.246] has joined ##hplusroadmap
13:53 -!- Malvolio [~Malvolio@unaffiliated/malvolio] has quit [Quit: Malvolio]
14:03 -!- deep-book-gk_ [~1wm_su@158.85.67.146] has joined ##hplusroadmap
14:05 -!- deep-book-gk_ [~1wm_su@158.85.67.146] has left ##hplusroadmap []
14:46 -!- augur [~augur@noisebridge130.static.monkeybrains.net] has quit [Remote host closed the connection]
14:50 -!- augur [~augur@noisebridge130.static.monkeybrains.net] has joined ##hplusroadmap
15:04 -!- CheckDavid [uid14990@gateway/web/irccloud.com/x-zkcevvguekhuyqtc] has quit [Quit: Connection closed for inactivity]
15:09 -!- Malvolio [~Malvolio@unaffiliated/malvolio] has joined ##hplusroadmap
15:30 -!- emeraldgreen [~user@92.222.68.248] has joined ##hplusroadmap
15:31 < emeraldgreen> kanzure nice comment https://news.ycombinator.com/item?id=14887559 exactly my thoughts. Costs of [life science] research are bloated 10-100x compared to optimized value, especially so in the west
15:31 < yoleaux> 28 Jul 2017 01:22Z <maaku> emeraldgreen: you mean like this? https://blockstream.com/2016/07/19/blockstream-defensive-patent-strategy.html
15:32 < emeraldgreen> kanzure If I could help this cause (lowering the costs of bio-research), I'd find time to do it
16:03 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has quit [Ping timeout: 260 seconds]
16:10 -!- entity8421[m] [entity8421@gateway/shell/matrix.org/x-vryklvubjdvmcnxq] has quit [Ping timeout: 276 seconds]
16:10 -!- Rmesil8O4b[m] [rmesil8o4b@gateway/shell/matrix.org/x-jzvffiykjtlqhmyo] has quit [Ping timeout: 264 seconds]
16:11 < gwillen> hey, I forget if I've asked this here before, but I'm curious what people here think of https://foreverlabs.com/
16:26 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has joined ##hplusroadmap
16:29 < kanzure> i think i asked them if they store chutzpah and they replied "no".
16:37 -!- entity8421[m] [entity8421@gateway/shell/matrix.org/x-sloyysuotfmmfxrw] has joined ##hplusroadmap
16:46 < kanzure> .tw https://twitter.com/bcrypt/status/891739574669459456
16:46 < yoleaux> chief medical officer of Intel on digital biology: "The human genome is full of potential exploits." #defcon https://pbs.twimg.com/media/DGAYZyNVYAAbkVN.jpg https://pbs.twimg.com/media/DGAYZyMVwAEP12F.jpg (@bcrypt)
16:47 -!- delinquentme [~delinquen@wsip-24-120-55-70.lv.lv.cox.net] has joined ##hplusroadmap
16:47 -!- Rmesil8O4b[m] [rmesil8o4b@gateway/shell/matrix.org/x-pxfnyjfjhvkngcln] has joined ##hplusroadmap
16:48 < kanzure> https://twitter.com/bcrypt/status/891752069811511296
16:49 < kanzure> tw
16:49 < kanzure> .tw
16:49 < yoleaux> "code signing to authenticate genome updates? what could possibly go wrong with having an x509 parser in your DNA" (@bcrypt, in reply to tw:891751001459310592)
16:50 -!- TinKode [~TinKode@unaffiliated/tinkode] has quit [Read error: Connection reset by peer]
16:53 < kanzure> here is one about increasing genetic susceptibility to extremely embarrassing odor and/or diarrhea https://twitter.com/triptananda/status/891803230119833602
16:53 < kanzure> .tw https://twitter.com/triptananda/status/891803230119833602
16:53 < yoleaux> @bcrypt Check the last one about involuntary obscenities - the white house press secretary's proteins have been hacked by Russia 🇷🇺 https://pbs.twimg.com/media/DGBSTdcXoAQBXO0.jpg (@triptananda, in reply to tw:891739574669459456)
16:57 -!- darsie [~darsie@84-113-55-42.cable.dynamic.surfer.at] has joined ##hplusroadmap
16:57 < kanzure> .tw https://twitter.com/FernRoyal/status/891775281689747457
16:57 < yoleaux> @bcrypt @BrianTRice There was an episode from the not so famous SF tv series Sliders where there was DNA money. You literally carried your money inside yourself (@FernRoyal, in reply to tw:891752069811511296)
16:58 < kanzure> ehh i guess you could put some private keys in dna... but why would you want to?
16:59 -!- TinKode [~TinKode@unaffiliated/tinkode] has joined ##hplusroadmap
17:03 < kanzure> "sunlight intolerance to perpetuate full veiling of women" uhhh...
17:29 < maaku> nmz787: azure is probably much cheaper for GPU tests
17:30 < maaku> at least that was true some 6-12 months ago. there's new AWS gpu instances I think
17:31 < kanzure> maaku: google has a tensorflow-specific service that either directly uses GPUs on demand or does some other clever thing with asics https://cloud.google.com/tpu/
17:31 < maaku> and putting private keys in the worlds most successful replicating molecule? sounds like a bad idea
17:32 < kanzure> right.
17:32 < maaku> kanzure: I thought TPUs only did inference? the site seems to imply both
17:33 < kanzure> it's generic tensorflow stuff.
17:33 < kanzure> machine learning computronium i guess.
17:37 < kanzure> maaku: so any chance of blockstream doing crypto-biology stuff?
17:40 < maaku> unlikely to get management support, although there's a lot of interest among individuals
17:40 < maaku> not just engineers. Alex did policy at 23andme
17:40 < maaku> kanzure: you have projects in mind?
17:43 < kanzure> error correction codes using some of these: http://diyhpl.us/wiki/gene-editing/
17:43 < kanzure> https://groups.google.com/d/msg/enzymaticsynthesis/WvDidIldm7c/q7aQTnmkAAAJ
17:44 < kanzure> also for your stamps i think you can encode extra information in dna in the liquid ink if you want
17:47 < nmz787> maaku: I'm trying to decide to get $600, $800, or $100 video card for a new computer I have to build around a 22 core Xeon
17:49 < maaku> nmz787: depends how much you are likely to actually use the card I guess, and with what RAM/model size requirements you have
17:49 < maaku> (if you're just starting out or playing around with small models to learn, get the cheapest card and upgrade later to the next generation)
17:50 < maaku> er, not the cheapest. the less expensive but actually workstation-quality
17:50 < nmz787> yeah that's kinda what I'm thinking, I don't game, but would like to watch 4k tv/videos without stuttering
17:50 < maaku> ah well the $100 card is fine for that
17:51 < maaku> actually better than a workstation card as the consumer GPUs have hardware video decode
17:52 < kanzure> x86 cpu fuzzer https://github.com/xoreaxeaxeax/sandsifter https://github.com/xoreaxeaxeax/sandsifter/blob/master/references/domas_breaking_the_x86_isa_wp.pdf
18:07 -!- deep-book-gk [~1wm_su@d8116.artnet.gda.pl] has joined ##hplusroadmap
18:09 -!- deep-book-gk [~1wm_su@d8116.artnet.gda.pl] has left ##hplusroadmap []
18:54 -!- delinquentme [~delinquen@wsip-24-120-55-70.lv.lv.cox.net] has quit [Ping timeout: 255 seconds]
19:19 -!- delinquentme [~delinquen@wsip-24-120-55-65.lv.lv.cox.net] has joined ##hplusroadmap
19:22 -!- emeraldgreen [~user@92.222.68.248] has quit [Ping timeout: 255 seconds]
19:32 -!- fltrz [4d6d6745@gateway/web/freenode/ip.77.109.103.69] has quit [Ping timeout: 260 seconds]
19:36 -!- emeraldgreen [~user@188.227.115.178] has joined ##hplusroadmap
20:02 -!- TinKode [~TinKode@unaffiliated/tinkode] has quit [Quit: TinKode]
20:09 -!- aeiousomething [~aeiousome@124.123.15.66] has joined ##hplusroadmap
20:31 -!- q4 [~q4@user-94-254-233-132.play-internet.pl] has joined ##hplusroadmap
20:57 < fenn> nmz787: k80 gpu was too slow for running a premade deepspeech network in real time
20:57 < fenn> you can at least use its benchmarks to decide what is adequate
20:57 < fenn> it ran at maybe 0.7x speed
21:00 -!- Gurkenglas__ [~Gurkengla@dslb-178-005-166-215.178.005.pools.vodafone-ip.de] has quit [Ping timeout: 276 seconds]
21:14 -!- q4 [~q4@user-94-254-233-132.play-internet.pl] has quit [Quit: Textual IRC Client: www.textualapp.com]
21:34 < kanzure> caching: i got a magical 80x speedup, but i don't remember where i got this from. i'll just proceed to assume it's borrowed from the future for now.
21:44 -!- aeiousomething [~aeiousome@124.123.15.66] has quit [Ping timeout: 246 seconds]
21:46 -!- aeiousomething [~aeiousome@124.123.15.66] has joined ##hplusroadmap
22:30 -!- GK1wmSU [~deep-book@83.143.245.6] has joined ##hplusroadmap
22:33 -!- GK1wmSU [~deep-book@83.143.245.6] has left ##hplusroadmap []
22:39 -!- yashgaroth [~yashgarot@2606:6000:cd4d:3300:f5e0:f867:a11d:8d52] has quit [Quit: Leaving]
22:43 -!- _GK1wmSU [~deep-book@d8116.artnet.gda.pl] has joined ##hplusroadmap
22:45 -!- _GK1wmSU [~deep-book@d8116.artnet.gda.pl] has left ##hplusroadmap []
23:13 -!- aeiousomething [~aeiousome@124.123.15.66] has quit [Ping timeout: 260 seconds]
23:13 -!- aeiousomething [~aeiousome@124.123.15.66] has joined ##hplusroadmap
23:31 < nmz787> fenn: hmm, some quick googling shows that the K80 is a memory-slower architecture than even the GTX 1080... so maybe this new Vega card which was just announced will work better... it is supposed to have even faster/more memory bandwidth
23:40 -!- augur [~augur@noisebridge130.static.monkeybrains.net] has quit [Remote host closed the connection]
23:40 -!- poppingtonic [~brian@unaffiliated/poppingtonic] has quit [Remote host closed the connection]
23:43 -!- fltrz [4d6d6745@gateway/web/freenode/ip.77.109.103.69] has joined ##hplusroadmap
23:43 < fltrz> that sandsifter paper is phenomenal!
23:45 < fltrz> tldr: by putting candidate instruction sequence on page boundaries (one page executable, other not), the page fault exception can leak the baremetal's interpretation of instruction end
23:45 < fltrz> shifting the byte sequence until there is no pagefault determines the length
23:46 < fltrz> this way even ring -2 (System Management Mode) instructions can be identified from ring 3 (user mode)
23:47 < mrdata> page fault is unreliable
23:48 < fltrz> mrdata: the explanation in the paper is a bit more nuanced of course, but I am curious what facet of page fault unreliability you are referring to?
23:49 -!- mrdata_ [~mrdata@unaffiliated/mrdata] has joined ##hplusroadmap
23:49 < mrdata_> how do you force a fault?
23:51 < fltrz> I only read the paper, they seem to execute the start of a 15-byte sequence, which starts in an executable page, and the processor will eventually end up trying to fetch from the non-executable page
23:51 < mrdata_> maybe i'll have to read the paper
23:52 < mrdata_> hmm
23:52 -!- mrdata [~mrdata@unaffiliated/mrdata] has quit [Ping timeout: 255 seconds]
23:53 < fltrz> I highly recommend it, but the explanation is a bit in reverse, they explain the depth first traversal *assuming we have a means of detecting instruction length* (during which I grew irritated because I didn't know any such means) and only in the next section describe how they detect instruction length
23:54 < fltrz> which at once resolved the tension.
23:55 < fltrz> its only 5 or 6 pages
23:56 < fltrz> so the fault is forced by trying to execute code residing in a non-executable page
--- Log closed Mon Jul 31 00:00:36 2017