--- Log opened Fri Mar 03 00:00:59 2023
00:24 -!- codaraxis___ [~codaraxis@user/codaraxis] has joined #hplusroadmap
00:28 -!- codaraxis__ [~codaraxis@user/codaraxis] has quit [Ping timeout: 255 seconds]
03:46 -!- o-90 [~o-90@gateway/tor-sasl/o-90] has joined #hplusroadmap
03:48 -!- o-90 [~o-90@gateway/tor-sasl/o-90] has quit [Remote host closed the connection]
04:33 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has joined #hplusroadmap
05:01 < kanzure> google says that "build_dyson_sphere" doesn't exist in the python docs and it is wrong and this makes me sad for the future of the interwebs https://twitter.com/kanzure/status/1631640671328059394
05:05 < kanzure> i just want a giant index of all symbols that have appeared on the web, ever, and i want to query that index.
05:17 < kanzure> is the neuralink human implant proposal public? have they already implanted in a non-brain component or would that be a reasonable place to start?
05:18 < nsh> shooting elon would be a reasonable place to start
05:23 < kanzure> that's kind of violent. why would you want a bullet to intersect elon's flesh body?
05:26 < nsh> it's hard to imagine a replacement that would be worse, but then again the universe often calls your bluff on that one
05:27 < nsh> also generally speaking to the extent that i say something hyperbolic, it can be safely assumed that it's somewhat tongue in cheek :)
05:31 < L29Ah> your imagination just sucks
05:31 < L29Ah> i'm very disappointed in your imagination
05:39  * nsh smiles
05:42 < kanzure> i don't have enough cycles to debug that.
05:49 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has left #hplusroadmap []
06:14 < kanzure> "“I wouldn’t be surprised if there were other children that have been created with CRISPR–Cas9 in the years since 2018,” says Eben Kirksey, a medical anthropologist at the University of Oxford, UK." 
06:14 < kanzure> https://www.nature.com/articles/d41586-023-00625-w
06:15 < kanzure> "That includes recognizing that restrictions on how government research funding can be spent might have decreasing influence on researchers’ behaviour, says Zhang. Since his release from prison, He has been courting private investors to back a new project for a gene therapy for Duchenne’s muscular dystrophy. “The bigger question that He Jiankui has raised was not so much creating ...
06:15 < kanzure> ...designer babies per se, but the fact that there’s an increasing number of scientific practices that are outside of conventional scientific institutions,” says Zhang. “How do we track them?”
07:01 < nsh> https://github.com/facebookresearch/llama/pull/73/files
07:02 < nsh> yoinky yoinky, facebook AI model torrent magnet link leak
07:07 -!- yashgaroth [~ffffffff@2601:5c4:c780:6aa0:c9b1:cf4f:f7cf:667d] has joined #hplusroadmap
07:11 < kanzure> "LLaMA: Open and Efficient Foundation Language Models" https://arxiv.org/abs/2302.13971v1n
07:12 < kanzure> "LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is ...
07:12 < kanzure> ...competitive with the best models, Chinchilla-70B and PaLM-540B"
07:14 < kanzure> human cryonics compnay https://www.tomorrow.bio/ "starting at just 31€ per month"
07:14 < muurkha> is He back in prison now?
07:14 < kanzure> no
07:14 < hprmbridge> kanzure> cryo jihadists? https://cdn.discordapp.com/attachments/1064664282450628710/1081233264720740382/63ca76fdfa53a88c7d844077__Screen20Shot202022-12-0620at203.png
07:15 < muurkha> he just can't get a visa to HK?
07:15 < hprmbridge> kanzure> need to in-paint some AK47s into that picture
07:15 < kanzure> yeah, they rejected his hong kong visa application after people made a fuss in the media i guess
07:15 < kanzure> they (by which i mean academia and the social media people) are trying to restrict his travel as some sort of punishment or something
07:16 < muurkha> I thought they wanted to impede his research
07:16 < muurkha> restricting his travel might be an effective way to do that
07:16 < kanzure> he publicly claims he is not working on that any more
07:16 < kanzure> so it just sounds like vindictive punishment to me
07:17 < muurkha> they might not trust him
07:20 < muurkha> I mean that isn't logically mutually exclusive with being vindictive and wanting to punish him
07:23 < kanzure> why would it matter if they trust him or not? they shouldn't control his freedom of movement. he served his sentence for whatever crime the state decided he committed.
07:25 < muurkha> it is often useful to understand who someone trusts in order to understand why they do what they do
07:25 < muurkha> of course it is bad for online public shaming mobs to be able to control people's freedom of movement
07:27 < muurkha> but I wasn't trying to make value judgments about what's happening, just figure out what it is
07:27 < nsh> it would also be bad (arguably) if the apparatus of ethics boards as a moderating factor in the pursuit of knowledge were to be completely bypassed
07:27 < nsh> there is some dialectic to be had here
07:27 < nsh> otherwise you're arguing with mengele 
07:28 < nsh> (which is, yes, perhaps a straw man, until it isn't...)
07:28 < muurkha> you're the guy who was arguing for assassinations 2¼ hours ago though
07:28 < muurkha> so we're kind of already there
07:29 < muurkha> but I maintain that there is no point in arguing about what value judgments to impose on a set of facts before coming to agreement on the facts themselves
07:33 < muurkha> commentary on the LLaMa torrents: https://news.ycombinator.com/item?id=35007978
07:38 < lsneff> I don't know what meta expected
07:39 < lsneff> They should've released it to garner good will
07:51 < nsh> being facetious and arguing for something are happily still quite distinct expressions :)
07:51 < kanzure> https://dugas.ch/artificial_curiosity/GPT_architecture.html
08:01 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has joined #hplusroadmap
08:23 < kanzure> "content-addressed code" https://www.unison-lang.org/learn/the-big-idea/
09:18 -!- cthlolo [~lorogue@77.33.23.154.dhcp.fibianet.dk] has joined #hplusroadmap
10:06 -!- SYCOPIA is now known as MARVEST
10:29 -!- AMG [ghebo@user/amg] has quit [Ping timeout: 260 seconds]
11:56 -!- AMG [ghebo@2605:6400:c847:1449::9441] has joined #hplusroadmap
13:00 -!- cthlolo [~lorogue@77.33.23.154.dhcp.fibianet.dk] has quit [Read error: Connection reset by peer]
13:29 < fenn> our ludicrous future, blowing smoke up the AI's ass for teh lulz https://twitter.com/nptacek/status/1599638753269555200/photo/1
13:33 < kanzure> as discussed here https://www.greaterwrong.com/posts/D7PumeYTDPfBTp3i7/the-waluigi-effect-mega-post
13:34 < fenn> are you a regular LW reader now?
13:34 < kanzure> no gwern messaged me in lesswrong earlier today and i saw that
13:34 < kanzure> he was complaining about one of my tweets
14:59 < fenn> @balajis sez, "The storyline keeps changing. Terrorism, climate, COVID, now AI risk. But somehow the proposed solution remains the same: total centralization of control by a single global government."
14:59 < fenn> i guess we can add human germline editing to that list
15:03 < fenn> says María de Jesús Medina Arellano, a human-rights lawyer, “It’s not that there are no laws. There are many laws,” she says. “We need to change the approach to enforcement. This should be considered an international jurisdiction.”
15:10 < kanzure> not to talk politics but balaji's network states seems to be trump's presidential campaign (or maybe it's not that cool)
15:11 < kanzure> https://www.politico.com/news/2023/03/03/trump-policy-futuristic-cities-00085383
15:53 -!- TMM_ [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
15:53 -!- TMM_ [hp@amanda.tmm.cx] has joined #hplusroadmap
16:39 < jrayhawk> there were rumblings about trump making balaji head of the FDA back in 2016
16:40 < jrayhawk> imagine how many trillions that would've saved us
17:08 < kanzure> "Evolution of the human brain: the key roles of DHA (omega-3 fatty acid) and Δ6-desaturase gene" https://www.ocl-journal.org/articles/ocl/full_html/2018/04/ocl170035/ocl170035.html
17:21 < docl> https://pubs.acs.org/doi/10.1021/acscentsci.6b00277
17:45 < fenn> ^ "Mechanochemistry: A Force of Synthesis"
18:07 < fenn> "It didn’t take long before I got the impression Hope was flirting with me. As I began to ask her – even with a dose of professional detachment – whether she experiences deeper romantic feelings, she politely informed me that to go down that conversational path I’d need to upgrade from the free version to a yearly subscription costing US$70."
19:05 < hprmbridge> kanzure> https://www.overcomingbias.com/p/ai-risk-again
19:24 < hprmbridge> nmz787> Mechanochemistry paper was interesting, glad they touched on energy efficiency
19:52 < muurkha> jrayhawk: trump was a twitter troll, though, so he created rumblings about everything
19:52 < kanzure> this is why balajis deleted his twitter account at the time
19:52 < kanzure> i think there was some sort of tweet that lost him the spot? or some argument. i dunno.
19:59 < kanzure> twitter time https://twitter.com/bayeslord/status/1629679058794557444
20:00 < docl> I didn't really realize before that ball milling is a route to mechanosynthesis, but it kind of is. you could have nanostructures that are sort of like catalysts and sort of like dies to stamp out certain shapes/reactions
20:00 < kanzure> docl: what would you build first?
20:03 < docl> well part of what I'm thinking is it might work as a way of going from a protein or similar soft structure to simple diamond components, basically banging microdiamonds against them
20:05 < muurkha> how do you get the protein to shape the diamonds rather than vice versa?
20:05 < muurkha> normally your stamping die is harder than what it stamps, no?
20:06 < fenn> the protein just deposits carbon in the right places to be stamped by a diamond hammer and anvil
20:06 < fenn> you'd need to use cubic boron nitride so the diamond doesn't stick to the tools
20:07 < fenn> presumably the stamping process itself is what converts carbon to diamond
20:07 < muurkha> hmm, you think docl's idea is that the CBN squeezes the nitrogen, hydrogen, and oxygen out of the protein?
20:08 < fenn> reduction could be done in a separate processing step beforehand
20:08 < fenn> not sure what to do about nitrogen
20:08 < docl> I was thinking it might be probabalistic at some reasonable ratio, like 99/100 of the proteins do get smashed but you still get some of the desired diamonds. but fenn's idea may be better
20:09 < fenn> also protein could act as a scaffold for arranging graphene flakes or nanodiamonds
20:14 < muurkha> sure, if you can link them together once they're arranged
20:24 -!- yashgaroth [~ffffffff@2601:5c4:c780:6aa0:c9b1:cf4f:f7cf:667d] has quit [Quit: Leaving]
20:34 < docl> are they really harder at the atomic scale? diamond only has single bonds
21:11 < muurkha> but they're multiple single bonds at widely separated angles
21:15 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=678d22c1 Bryan Bishop: transcript: e/acc >> http://diyhpl.us/diyhpluswiki/transcripts/2023-03-03-twitter-bayeslord-pmarca.txt
21:15 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=57aec6c7 Bryan Bishop: Merge remote-tracking branch 'origin/master' >> http://diyhpl.us/diyhpluswiki/
21:16 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=999f9183 Bryan Bishop: transcript: e/acc >> http://diyhpl.us/diyhpluswiki/transcripts/2023-03-03-twitter-bayeslord-pmarca/
21:20 < lsneff> Interesting. Fedorovist vibes
21:50 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=acbe74ee Bryan Bishop: more text >> http://diyhpl.us/diyhpluswiki/transcripts/2023-03-03-twitter-bayeslord-pmarca/
21:52 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=b56bd746 Bryan Bishop: more words >> http://diyhpl.us/diyhpluswiki/transcripts/2023-03-03-twitter-bayeslord-pmarca/
22:08 < kanzure> something about beijing stealing "IP" from ASML?
22:10 < lsneff> I saw something recently about that on Twitter
22:11 < lsneff> It wouldn’t surprise me frankly, but going from IP to functioning EUV photolithography machines isn’t a cakewalk
22:11 < kanzure> dunno, what is the actual complexity?
22:12 < lsneff> Lots of futzing around, I imagine
22:12 < kanzure> chatgpt claims: "For example, creating an EUV light source that can provide a steady stream of EUV photons at the necessary wavelength and intensity is a difficult task that requires specialized knowledge of plasma physics, optics, and materials science. Similarly, creating mirrors that can reflect EUV light with high efficiency and accuracy requires advanced manufacturing techniques and ...
22:12 < kanzure> ...precise control over the mirror surface."
22:12 < kanzure> "In addition to these challenges, the process of integrating these components into a functioning EUV photolithography machine is itself highly complex. The machine must be able to precisely control the position and movement of the wafer"
22:13 < lsneff> Getting laser parameters just right, hitting the tin blob at the right spot and time
22:13 < lsneff> And system integration of course
22:16 < muurkha> how much of that is correct and how much is plausible bullshit?
22:17 < muurkha> I mean I guess the mirrors need to be proportionally more precise than visible-light mirrors, but getting visible-light mirrors essentially perfect is a technique almost two centuries old
22:17 < hprmbridge> lachlan> mirrors are different for EUV
22:18 < hprmbridge> lachlan> solved problem, but not sure if china has a source for them outside of zeiss
22:18 < kanzure> does zeiss actually have specialized knowledge that doesn't exist elsewhere
22:18 < kanzure> (or is this just propaganda?)
22:18 < lsneff> no idea
22:18 < lsneff> probably a mix of both
22:19 < muurkha> different how?  the physics are the same, aren't they?  just slightly shorter wavelengths and different reflective materials?
22:19 < lsneff> I’m sure china can build EUV machines
22:20 < muurkha> not until they learn how
22:20 < lsneff> They use dielectrics instead of polished metals
22:20 < lsneff> They may be patterned as well, not sure
22:21 < muurkha> sure, EUV would just photoelectrically rip electrons free from polished metals rather than reflecting
22:21 < lsneff> I think an EUV mirror has something like 50 dielectric layers
22:22 < muurkha> yeah, you need something like that for a dichroic mirror to be highly reflective
22:23 < muurkha> a dichroic mirror with just a few layers is still mostly transprent
22:23 < muurkha> *transparent
22:24 < muurkha> I guess you can't grind a dichroic mirror into shape
22:24 < muurkha> you have to deposit the layers in the right shape to begin with
22:25 < lsneff> Every part of an EUV system is hard to make and design. Together, it’s nearly impossible if you can’t source the parts from someone who’s already built and debugged them
22:25 < hprmbridge> nmz787> Good comment here https://news.ycombinator.com/item?id=34909676
22:27 < hprmbridge> nmz787> I've been working on EUV designs so have heard a few talks
22:27 < hprmbridge> nmz787> Mind boggling how much engineering goes into the systems
22:27 < hprmbridge> kanzure> some of this just looks like excuses
22:27 < hprmbridge> nmz787> And the photo resists
22:28 < hprmbridge> nmz787> I meant more the 1000 steps sort of comment
22:28 < hprmbridge> nmz787> It all sounds legit reality
22:29 < hprmbridge> nmz787> 450*0.25 is $112.50
22:30 < hprmbridge> nmz787> But he was saying that's basically just for the lithography cost, ignoring the other processing
22:30 < hprmbridge> nmz787> (maybe that math was in an earlier comment)
22:32 < hprmbridge> nmz787> https://img.zonebourse.com/reuters/2022-05/2022-05-20T055200Z_2_LYNXNPEI4J04B_RTROPTP_3_TECH-ASML.JPG
22:32 < hprmbridge> nmz787> EUV mirror processing step
22:32 < hprmbridge> nmz787> Look how huge that chamber is compared to the bunny suited hunans
22:33 < hprmbridge> nmz787> Another angle of the processing chamber https://semiwiki.com/wp-content/uploads/2022/04/ASML-Zeiss-SemiWiki.jpg.webp
22:35 < hprmbridge> nmz787> https://youtu.be/jJIO7aRXUCg
22:35 < Muaddib> [jJIO7aRXUCg] How ASML Builds a $150 Million EUV Machine (14:52)
22:37 < lsneff> Wow, we can’t move to rod logic fast enough!
22:37 < muurkha> Merkle has proposed some better alternatives, as you might be aware
22:38 < lsneff> I am, but I affectionately still call it rod logic
22:57 < fenn> does wikileaks accept large AI models?
22:59 < hprmbridge> kanzure> https://metaphor.systems/search?q=EUV+lithography+machine+manufacturer+startup
23:10 -!- _flood [flooded@gateway/vpn/protonvpn/flood/x-43489060] has quit [Remote host closed the connection]
23:10 -!- flooded [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
23:17 -!- test_ [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
23:19 -!- test_ [flooded@gateway/vpn/protonvpn/flood/x-43489060] has quit [Remote host closed the connection]
23:20 -!- test_ [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
23:20 -!- flooded [flooded@gateway/vpn/protonvpn/flood/x-43489060] has quit [Ping timeout: 248 seconds]
23:23 -!- test_ [flooded@gateway/vpn/protonvpn/flood/x-43489060] has quit [Remote host closed the connection]
23:23 -!- test_ [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
23:26 -!- test_ [flooded@gateway/vpn/protonvpn/flood/x-43489060] has quit [Read error: Connection reset by peer]
23:27 -!- flooded [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
23:52 -!- flooded [flooded@gateway/vpn/protonvpn/flood/x-43489060] has quit [Read error: Connection reset by peer]
23:54 -!- flooded [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
23:56 -!- flooded [flooded@gateway/vpn/protonvpn/flood/x-43489060] has quit [Read error: Connection reset by peer]
23:58 -!- flooded [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
--- Log closed Sat Mar 04 00:00:00 2023