--- Log opened Sun Jun 16 00:00:43 2019
--- Day changed Sun Jun 16 2019
00:00 -!- LeoTal1 [~Adium@118.170.86.79.rev.sfr.net] has joined ##hplusroadmap
00:03 -!- LeoTal [~Adium@28.170.86.79.rev.sfr.net] has quit [Ping timeout: 258 seconds]
02:02 -!- mrdata is now known as mrdata-
02:11 -!- mrdata- is now known as mrdata
02:15 -!- mrdata is now known as mrdata-
03:55 -!- mauz555 [~mauz555@2a01:e35:8ab1:dea0:c598:f363:e9a3:b040] has joined ##hplusroadmap
04:29 -!- transhumanist [~transhuma@2601:196:8800:6719:213a:525d:ab76:c7fe] has quit [Quit: Leaving]
04:55 -!- mauz555 [~mauz555@2a01:e35:8ab1:dea0:c598:f363:e9a3:b040] has quit [Remote host closed the connection]
05:07 -!- mauz555 [~mauz555@2a01:e35:8ab1:dea0:c598:f363:e9a3:b040] has joined ##hplusroadmap
05:28 -!- preview [~quassel@2407:7000:8423:b75:5983:c203:baaa:a2a0] has quit [Ping timeout: 252 seconds]
06:01 < kanzure> .tw https://twitter.com/KennethHayworth/status/1139698130172219392
06:01 < yoleaux> "Complete nanometer-level reconstruction of an entire mouse brain is a daunting prospect, as it requires detailed EM imaging of roughly 300,000 brain slices... Accomplishing such ambitious goal in a 5-year period will require continued advances in... https://twitter.com/AdrianAWanner1/status/1139675553630425100 (@KennethHayworth)
06:01 < kanzure> .tw https://twitter.com/KennethHayworth/status/1139698131451334656
06:01 < yoleaux> high-throughput tissue handling, EM devices with nearly 100 parallel beams, and modern machine learning algorithms implemented by supercomputers. Despite the enormity of the challenge, we believe it is feasible." https://acd.od.nih.gov/documents/reports/06142019BRAINReport.pdf (@KennethHayworth, in reply to tw:1139698130172219392)
06:01 < kanzure> .tw https://twitter.com/kanzure/status/1140242713826209792
06:01 < yoleaux> @KennethHayworth Hi Ken & @neurowitz; I vaguely recall someone talking about microchips that produce 100 parallel SEM beams and does scanning. Do you have any references on this? Who was doing this? (@kanzure, in reply to tw:1139698131451334656)
06:03 < kanzure> neurowitz is arguing that they should focus on multi-region neuronal physiology studies instead of just connectivity/connectomics.
06:03 < kanzure> for deep brain tissue the only way i see that being feasible is organoids and tissue culture, i don't know about the reliability of identifying individual target neurons from sliced brain and culturing just those neurons...
06:04 < kanzure> ken is also talking about memory decoding experiments, but between him and mcintyre i haven't heard any specifics beyond proving one or two bits of contextual fear memory-induced physical/physiological differences
06:05 -!- mauz555 [~mauz555@2a01:e35:8ab1:dea0:c598:f363:e9a3:b040] has quit [Remote host closed the connection]
06:05 < kanzure> in fact, i haven't heard any reasonable speculation about proving the presence of, or decoding, multiple bits of memory from brain tissue
06:34 < kanzure> what was the reference for the technique of converting a basic C program into a specific neural network that does the same computation?
06:37 < kanzure> here's an overkill method for memory decoding from brain tissue: using the cellular recording concept, make a set of different molecular events to record (mRNA transcript levels, protein levels, protein-protein interactions, neurotransmitter levels, action potentials, etc). then use identically structured neuronal networks in tissue culture and, optionally, do real-time patch clamp recording ...
06:37 < kanzure> ...and sampling.
06:37 < kanzure> you'd train the network to do basic arithmetic or pulse out a certain sequence
06:37 < kanzure> and then you'd collect all of this recorded data (via dna sequencing) and throw it into your magic machine learning device to figure out if any of the molecular events in some combination reliably correspond to decoding some number of bits from the tissue
06:38 -!- mauz555 [~mauz555@2a01:e35:8ab1:dea0:c598:f363:e9a3:b040] has joined ##hplusroadmap
06:39 < kanzure> everyone wants something simple like a direct-read method along the lines of "just read the connectivity and get the memories from that"
06:39 < kanzure> but it could be more like a direct read of (sample these 10 proteins and their distributions, these mRNA transcripts, the relative distribution of receptors on each synapse in this neuron)
06:42 < kanzure> actually you might be able to get away with something more primitive
06:43 < kanzure> say you have an 8-cell network and you're doing patch clamp on every single cell in the network; if you precisely sample each neuron every few hundred microseconds via patch clamp and store those samples isolated from each other in a retrievable way, then you could do conventional lab testing of each sample instead of that convoluted cellular recording DNA idea.
06:44 < kanzure> i suppose this assumes that the data you want is based on pH levels or whatever's floating around, rather than structural features like embedded receptor distribution in synapses
06:50 -!- mauz555 [~mauz555@2a01:e35:8ab1:dea0:c598:f363:e9a3:b040] has quit []
06:52 < kanzure> .tw https://twitter.com/KennethHayworth/status/1139304910707744769
06:52 < yoleaux> Spiny ‘memory’ synapses surely contain complex molecular state machines implementing learning rules. But information that a single synapse provides during the 100’s of ms required to recognize an object or recall a memory must be much simpler –size of EPSC (synaptic weight)... (@KennethHayworth)
06:52 < kanzure> .tw https://twitter.com/KennethHayworth/status/1139304912045694977
06:52 < yoleaux> If you can recall a memory in 100’s of ms then it is encoded at the synaptic weight level, not at the local state machine level, epigenome level etc. Neuronal dynamics at that timescale is defined by synaptic connectivity and by the receptors and ion channels already in place... (@KennethHayworth, in reply to tw:1139304910707744769)
06:55 < kanzure> .tw https://twitter.com/KennethHayworth/status/1123732478399795202
06:55 < yoleaux> These artificial memory experiments (also https://www.sciencedirect.com/science/article/pii/S2211124715002703) use paired photostimulation of two ensembles to encode a memory by strengthening interposed synapses. Are there experiments that attempt to decode a memory instead? #neuroscience (@KennethHayworth, in reply to tw:1123728794244796417)
06:55 < kanzure> "Artificial association of pre-stored information to generate a qualitatively new memory" https://www.sciencedirect.com/science/article/pii/S2211124715002703
06:57 < kanzure> "Memory formation in the absence of experience" https://www.nature.com/articles/s41593-019-0389-0 https://twitter.com/Franklandlab/status/1122898899906310144 https://twitter.com/KennethHayworth/status/1123728794244796417
06:58 < kanzure> "Labelling and optical erasure of synaptic memory traces in the motor cortex" https://www.nature.com/articles/nature15257 "Hayashi-Takagi et al. 2015 labeled only those synapses in motor cortex that grew in size while learning a motor task, and then optically shrunk them selectively erasing that memory" https://twitter.com/KennethHayworth/status/1113993479754080256
07:27 -!- Urchin [~urchin@unaffiliated/urchin] has quit [Remote host closed the connection]
07:29 < kanzure> "RAMbleed: Reading bits in memory without accessing them" https://rambleed.com/docs/20190603-rambleed-web.pdf (rowhammer stuff)
07:48 -!- NikopolSohru [~NikopolSo@185.217.171.16] has joined ##hplusroadmap
07:48 -!- NikopolSohru [~NikopolSo@185.217.171.16] has quit [Changing host]
07:48 -!- NikopolSohru [~NikopolSo@unaffiliated/nikopolsohru] has joined ##hplusroadmap
07:55 -!- yashgaroth [~ffffffff@2606:6000:ca81:3100:958:4328:b109:e71b] has joined ##hplusroadmap
08:45 -!- LeoTal1 [~Adium@118.170.86.79.rev.sfr.net] has quit [Quit: Leaving.]
09:04 < fenn> parallel SEM chips was halcyon molecular
09:07 -!- LeoTal [~Adium@118.170.86.79.rev.sfr.net] has joined ##hplusroadmap
09:16 < kanzure> thanks
09:55 < nmz787> I don't think I ever read about parallel e-beam stuff re halcyon, it was always in the context of lithography and not imaging
09:56 < kanzure> because you need super fast lithography?
09:58 -!- mrdata- is now known as mrdata
10:01 < nmz787> yeah, an alternative to the crazy extreme UV stuff and multi-patterning
10:01 < nmz787> .wik extreme uv
10:01 < yoleaux> "Extreme ultraviolet radiation (EUV or XUV) or high-energy ultraviolet radiation is electromagnetic radiation in the part of the electromagnetic spectrum spanning wavelengths from 124 nm down to 10 nm, and therefore (by the Planck–Einstein equation) having photons with energies from 10 eV up to 124 eV (corresponding to 124 nm to 10 nm  …" — https://en.wikipedia.org/wiki/Extreme_UV
10:05 < nmz787> re EUV machines "Each unit has a starting price of $119M"
10:38 < fenn> so a 7 nm process is "super duper extreme UV"?
10:44 < nmz787> apparently <10nm is x-ray
10:45 < nmz787> but 10nm is the minimum feature size, not necessarily the minimum space between features
10:45 < nmz787> so you might pattern two 10nm features, with 7nm between them, for example
10:45 < fenn> how much semiconductor manufacturing actually happens in austin? or is it all design work?
10:45 < kanzure> there's a few facilities for cmos
10:46 < nmz787> from an intel perspective, I know only of validation work, but I'm in validation so I have limited perspective
10:46 < kanzure> and samsung
10:46 < fenn> supposedly samsung invested $1B recently, but who knows what that means
10:47 < nmz787> it means about 9 EUV machines
10:48 < fenn> lol
10:49 < fenn> i guess i'm just wondering where they actually make all this stuff
10:51 < nmz787> intel has a lot of manufacturing in phoenix AZ, r&d fabs around here in OR, and a lot of production in china, malaysia, and some in ireland
10:51 < nmz787> a small amount in NM
10:52 < nmz787> there's a bunch of other manufactories around here, Maxim, ON semi (I think).... some stuff across the river in Washington
10:53 < nmz787> and obv a bunch of manufacturing in the bay area (Sili Valley)
10:54 < fenn> why do they do semiconductor production in china and malaysia?
11:00 < fenn> this shows only one specialized fab in china: https://en.wikipedia.org/wiki/List_of_Intel_manufacturing_sites
11:03 -!- priontology [~scionce@66.205.193.168] has quit [Read error: Connection reset by peer]
11:08 -!- priontology [~scionce@66.205.193.168] has joined ##hplusroadmap
11:17 -!- priontology [~scionce@66.205.193.168] has quit [Read error: Connection reset by peer]
11:18 < jrayhawk> intel also has a fab in Israel
11:19 < fenn> several of them it seems
11:19 < fenn> i know intel is particularly committed to keeping jobs in the US
11:20 < fenn> i just don't understand why it would be an advantage to do semiconductor manufacturing in china vs where your designers and process researchers live
11:20 < fenn> it is sort of energy intensive but china isn't any cheaper than the US in that regard
11:23 -!- priontology [~scionce@66.205.193.168] has joined ##hplusroadmap
11:24 < jrayhawk> the processes are pretty well sorted by the time manufacturing gets rolling; development happens at D1X and D1D in Hillsboro where most of the designers and process researchers live
11:25 < jrayhawk> in the case of ireland, it's almost certainly for tax reasons, and china for reasons of materials and labor costs
11:26 -!- purpleshift [purpleshif@gateway/vpn/mullvad/purpleshift] has joined ##hplusroadmap
11:26 < jrayhawk> and selling into the chinese market is generally a pain unless it's locally manufactured
11:37 -!- spaceangel [~spaceange@ip-62-24-76-6.net.upcbroadband.cz] has joined ##hplusroadmap
11:46 < fenn> "human-embodied technology transfer"
11:46 < fenn> never underestimate the bandwidth of a station wagon full of foreign-educated engineers!
12:31 -!- LeoTal [~Adium@118.170.86.79.rev.sfr.net] has quit [Ping timeout: 245 seconds]
12:39 -!- NikopolSohru [~NikopolSo@unaffiliated/nikopolsohru] has quit [Remote host closed the connection]
12:39 -!- purpleshift [purpleshif@gateway/vpn/mullvad/purpleshift] has quit [Quit: .]
12:40 -!- NikopolSohru [~NikopolSo@185.217.171.16] has joined ##hplusroadmap
12:40 -!- NikopolSohru [~NikopolSo@185.217.171.16] has quit [Remote host closed the connection]
12:41 -!- NikopolSohru [~NikopolSo@185.217.171.16] has joined ##hplusroadmap
13:15 -!- justanotheruser [justanothe@gateway/vpn/nordvpn/justanotheruser] has quit [Quit: WeeChat 2.4]
13:29 -!- justanotheruser [justanothe@gateway/vpn/nordvpn/justanotheruser] has joined ##hplusroadmap
13:53 -!- hrofu [~hrofu@unaffiliated/hrofu] has quit [Ping timeout: 252 seconds]
13:55 -!- hrofu [~hrofu@unaffiliated/hrofu] has joined ##hplusroadmap
14:40 -!- CryptoDavid [uid14990@gateway/web/irccloud.com/x-wlcogierdmdavrak] has joined ##hplusroadmap
14:42 -!- hehelleshin [~talinck@98.29.27.253] has joined ##hplusroadmap
14:46 -!- helleshin [~talinck@98.29.27.253] has quit [Ping timeout: 248 seconds]
14:56 -!- priontology [~scionce@66.205.193.168] has quit [Ping timeout: 248 seconds]
15:35 -!- yashgaroth [~ffffffff@2606:6000:ca81:3100:958:4328:b109:e71b] has quit [Read error: Connection reset by peer]
15:37 -!- yashgaroth [~ffffffff@cpe-76-176-31-235.san.res.rr.com] has joined ##hplusroadmap
15:42 -!- justanotheruser [justanothe@gateway/vpn/nordvpn/justanotheruser] has quit [Ping timeout: 245 seconds]
15:43 -!- justanotheruser [justanothe@gateway/vpn/nordvpn/justanotheruser] has joined ##hplusroadmap
16:13 -!- cluelessperson [~cluelessp@unaffiliated/cluelessperson] has quit [Quit: Laters]
16:15 -!- cluelessperson [~cluelessp@unaffiliated/cluelessperson] has joined ##hplusroadmap
16:29 -!- NikopolSohru [~NikopolSo@185.217.171.16] has quit [Quit: Leaving]
16:40 -!- darsie [~kvirc@84-114-73-160.cable.dynamic.surfer.at] has quit [Ping timeout: 248 seconds]
17:02 -!- Urchin[emacs] [~user@unaffiliated/urchin] has joined ##hplusroadmap
17:33 -!- preview [~quassel@2407:7000:8423:b75:5983:c203:baaa:a2a0] has joined ##hplusroadmap
17:47 -!- CryptoDavid [uid14990@gateway/web/irccloud.com/x-wlcogierdmdavrak] has quit [Quit: Connection closed for inactivity]
17:48 < kanzure> .tw https://twitter.com/KennethHayworth/status/1139971474356355072
17:48 < yoleaux> If @NIH refuses to debate this possibility, they are telling terminal patients their only option is unregulated ‘back alley’ cryonics. If neuroscientists refuse to debate are they saying that even they don’t believe neuroscience will ever be successful? #BrainPreservationDebate (@KennethHayworth, in reply to tw:1139971472666038273)
17:48 < kanzure> .tw https://twitter.com/kanzure/status/1140255421527146496
17:48 < yoleaux> @KennethHayworth @NIH It doesn't need to be unregulated back alley cryonics. In these situations, it is very common for groups to form their own self-regulatory organizations and propose their own standards and rules. This is perfectly normal, and we have no right to force others to regulate it. (@kanzure, in reply to tw:1139971474356355072)
17:48 < kanzure> .tw https://twitter.com/KennethHayworth/status/1140357333240991744
17:48 < yoleaux> @kanzure @NIH Cryonics avoided regulation for decades by operating only on the legally dead. This allowed them to botch case after case with no repercussions, and avoid requirement for scientific publications etc. If Minsky could not get a decent preservation what makes you think you will? (@KennethHayworth, in reply to tw:1140255421527146496)
17:49 < kanzure> ken is either being an idiot or he's illiterate
18:10 -!- l_wl [~l_wl@ip-173-120-184-230.ekrgmd.spcsdns.net] has joined ##hplusroadmap
18:23 < jrayhawk> lol what? alcor and cryinst publish literally every case
18:24 < fenn> and in a significant fraction of them they screw up
18:25 < fenn> hayworth's position seems to be, that if they were treated as living patients then the success rate would be higher, because regulation magic
18:27 < jrayhawk> i am trying to understand what the hell he would want for publications; what regulated medical organization matches or exceeds that level of public disclosure?
18:27 < fenn> by scientific publications i think he means something like clinical trials to prove safety and efficacy of the procedures
18:28 < fenn> (and look how well that worked for the rest of healthcare)
18:29 < fenn> BANNED BY THE FDA - a lone scientist's struggle for recognition in the face of an oppressive ontology
18:29 < fenn> the latest harlequin paperback
19:22 -!- priontology [~scionce@75-169-153-91.slkc.qwest.net] has joined ##hplusroadmap
19:24 < kanzure> http://www.decisionproblem.com/paperclips/index2.html
19:38 -!- l_wl_ [~l_wl@ip-173-120-184-230.ekrgmd.spcsdns.net] has joined ##hplusroadmap
19:38 -!- l_wl [~l_wl@ip-173-120-184-230.ekrgmd.spcsdns.net] has quit [Read error: Connection reset by peer]
20:34 -!- l_wl_ [~l_wl@ip-173-120-184-230.ekrgmd.spcsdns.net] has quit []
20:49 -!- preview [~quassel@2407:7000:8423:b75:5983:c203:baaa:a2a0] has quit [Ping timeout: 248 seconds]
20:54 -!- preview [~quassel@2407:7000:8423:b75:5983:c203:baaa:a2a0] has joined ##hplusroadmap
21:29 -!- priontology [~scionce@75-169-153-91.slkc.qwest.net] has quit [Ping timeout: 244 seconds]
21:32 -!- yashgaroth [~ffffffff@cpe-76-176-31-235.san.res.rr.com] has quit [Quit: Leaving]
22:09 -!- spaceangel [~spaceange@ip-62-24-76-6.net.upcbroadband.cz] has quit [Ping timeout: 245 seconds]
23:14 -!- preview [~quassel@2407:7000:8423:b75:5983:c203:baaa:a2a0] has quit [Ping timeout: 248 seconds]
23:37 -!- priontology [~scionce@66.205.193.168] has joined ##hplusroadmap
23:42 -!- darsie [~kvirc@84-114-73-160.cable.dynamic.surfer.at] has joined ##hplusroadmap
--- Log closed Mon Jun 17 00:00:44 2019