--- Log opened Wed Feb 26 00:00:29 2025
01:26 < hprmbridge> fodagut> alethkit: need context. What do you want to do in nanotech? There’s a general need for synthetic chemists, applied physicists, surface science, chemical simulation (molecular mechanics and DFT), mechanical engineering, quantum sensing, control theory, and plenty of software dev.
03:39 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has joined #hplusroadmap
04:08 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has joined #hplusroadmap
04:13 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has left #hplusroadmap []
04:17 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has joined #hplusroadmap
04:27 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has left #hplusroadmap []
04:30 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has joined #hplusroadmap
04:47 < hprmbridge> kanzure> "Nanorobotic actuator based on interlayer sliding ferroelectricity and field-tunable friction" https://arxiv.org/abs/2408.16519
04:47 < hprmbridge> kanzure> blebbisomes https://www.nature.com/articles/s41556-025-01621-0
06:34 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has quit [Read error: Connection reset by peer]
07:19 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has joined #hplusroadmap
07:32 < alethkit> fodagut: My background is in PL dev/design, so probably either software dev in general, or chemical simulation/design/optimisation.
07:33 < hprmbridge> nmz787> PL??? Hopefully not Perl
07:46 < L29Ah> PL/SQL
09:41 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has quit [Remote host closed the connection]
09:41 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has joined #hplusroadmap
10:08 -!- etc-vi [~etc-vi@user/meow/girlchunks] has quit [Ping timeout: 272 seconds]
10:11 -!- etc-vi [~etc-vi@user/meow/girlchunks] has joined #hplusroadmap
10:27 < hprmbridge> kanzure> https://github.com/Wan-Video/Wan2.1
10:39 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has quit [Read error: Connection reset by peer]
10:40 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has joined #hplusroadmap
11:33 < hprmbridge> Eli> 23andme is getting into the blood-based-age (BBA) game. Their blood testing is $100 for the same tests I can get at MarekDiagnostics for $50. I'm wondering if I could just open source the algorithm they used to allow people to come up with their own BBA? They are basing their BBA off a public paper. https://blog.23andme.com/articles/introducing-biological-age-feature-for-23andme-total-health-
11:33 < hprmbridge> Eli> members
11:38 < hprmbridge> Eli> As a side thought, it's kind of interesting that the BBA is becoming popular. It seems to be a lot more accurate at predicting mortality than chronological age.
12:06 < kanzure> are there any LLM benchmark leaderboards that measure value/second in terms of dollar values generated? or some other way of estimating the actual value per token or per second.
12:07 < kanzure> (solving word problems is not necessarily an indicator of real-world value)
13:04 -!- NewtonTrendy [Newton@user/bopqod] has quit [Quit: WeeChat 3.5]
15:34 -!- TMM [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
15:34 -!- TMM [hp@amanda.tmm.cx] has joined #hplusroadmap
16:45 < hprmbridge> fodagut> alethkit: while programming-language-for-nanotech is certinly on topic here, we have separate discord where some of that discussion resides, and people currently working on the software side: https://discord.gg/2EdR6WArXc
17:03 < L29Ah> are there any human benchmark leaderboards that measure value/second in terms of dollar values generated?
17:07 < hprmbridge> kanzure> ceo pay?
17:21 < L29Ah> if ceo pay is the benchmark for humans, then hardware electricity and maintenance costs are the benchmarks for LLMs
17:39 < fenn> kanzure: it can be derived from tok/s and https://arxiv.org/abs/2502.12115
17:39 < fenn> .t
17:39 < saxo> [2502.12115] SWE-Lancer: Can Frontier LLMs Earn $1 Million from Real-World Freelance Software Engineering?
17:40 < fenn> there is some concern that the github issues dataset used for these benchmarks often contained the solutions
17:40 < fenn> uh requests/s not tok/s
17:43 < fenn> ah this is upwork not github, ignore that as irrelevant
17:54 < hprmbridge> kanzure> software work is notoriously hard to price and valud
17:54 < hprmbridge> kanzure> er, value
17:54 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has quit [Ping timeout: 252 seconds]
18:00 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has quit [Read error: Connection reset by peer]
18:03 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has joined #hplusroadmap
18:26 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has quit [Read error: Connection reset by peer]
19:30 -!- flyback [~flyback@2601:540:c701:900:bace:ea78:c7a6:5f85] has quit [Ping timeout: 245 seconds]
19:43 -!- flyback [~flyback@2601:540:c701:900:d221:5962:6cc4:8826] has joined #hplusroadmap
20:46 < RangerMauve> I think it's unreasonable to look at CEO pay + machine intelligence because it's not like CEOs are the smartest people in a company. They happen to have access to wealth and social clout and sometimes they manage not to squander that opportunity. A chatbot would need to build a career with social relationships before making decisions to get a fair comparison
20:47 < superkuh> scifi tangent: https://glyphpress.com/talk/2014/feral
20:48 < RangerMauve> An aside, I've been making progress on the EEG neural network classification stuff with some friends. Gonna do a blink detection model (as an easy win) and get more in the weeds of how to use transformers for this stuff
20:50 < RangerMauve> superkuh: TY for sharing
20:56 < hprmbridge> fodagut> RangerMauve: the question was abou value delivered per second, not intelligence; value and intelligence are at best only loosely correlated
20:57 < RangerMauve> Fair. I think it'd be hard to derive comparable value. Like, the LLM can "make decisions" or "talk to stakeholders" but it would need to be embedded in a social context to be able to compare value
20:58 < RangerMauve> I think I'm just  getting at it not being just the model bringing value but the environment it's in
20:59 < RangerMauve> I feel like unaccountable LLM based CEOs beholden to the usual vulture shareholders is going to be an even bigger hellhole than we're already in
21:36 < jrayhawk> That is a key word you just used, but not for the reasons you think.
21:36 < jrayhawk> The CEO executes the strategy adopted by the board. The board is paying for the privilege of externalizing accountability onto the CEO in the event of bad outcomes that might result from executing the strategy.
21:36 < RangerMauve> Could you elaborate?
21:38 < jrayhawk> In the event of legal, financial, or social failures, all injured parties are set up to blame the CEO and not the board.
21:38 < jrayhawk> This enables the board to fire the CEO and burn the CEO's reputation to the ground at any time.
21:39 < fenn> in olden times we just used a goat
21:39 < jrayhawk> https://en.wikipedia.org/wiki/Golden_parachute
21:40 < RangerMauve> Yeah. In the case of "machines" going wrong there's often this cycle of "Oh we didn't do anything bad, it was $VENDOR_NAME" and the vendor is like "Oh it wasn't our tool, $CLIENT_COMPANY used it wrong"
21:40 < RangerMauve> A lot of the negatives seem to magically dissipate in this loop and accountability gets ignored
21:41 < RangerMauve> I forget the specific examples offhand but there's been stuff to do with machine learning based systems in the last few years along these lines IIRC
21:42 < jrayhawk> Accountability diffusion is an effective approach, but it's not *as* good as publicly sacrificing the designated asshole.
21:42 < RangerMauve> Like a ritual sacrifice but this time it's just a demon whose seed you shuffle and name you regenerate without needing to deal with grooming another human for it
21:42 < RangerMauve> Hopefully
21:43 < RangerMauve> I hope there's humans that can die or retire in the loop in places of authority in the future
21:45 < jrayhawk> Of course, boards have to be careful that they don't pay more than the reputational damage the accountability internalization is worth... https://en.wikipedia.org/wiki/Stephen_Elop#CEO_of_Nokia
21:46 < fenn> RangerMauve: see also Best Korea
22:00 < jrayhawk> https://youtu.be/bVzi4fDzxq0 CharaChord finding uses in being slightly less unfashionable than Twiddlers.
22:21 < fenn> https://fennetic.net/mirrors/homepage2.nifty.com/perky/belt.htm
22:22 < fenn> sorry about the character encoding gibberish
22:22 < fenn> older browsers could auto detect the encoding i guess
22:30 < hprmbridge> Katylase> hi friends! how are you doing today?
22:36 < fenn> 25% more mopey and disoriented than usual
22:37 -!- flooded [flooded@gateway/vpn/protonvpn/flood/x-43489060] has quit [Ping timeout: 245 seconds]
22:39 < hprmbridge> Katylase> can I show you something? it might make your mood better^^
23:00 < fenn> ... of course
23:02 < fenn> she is a thing of beauty: https://github.com/deepseek-ai/DualPipe
23:05 < jrayhawk> years of balancing factorio belts pays off
23:05 < fenn> pretty much
23:07 < fenn> still a 100% useless website http://www.jblake.org/
23:10 < fenn> since i haven't complained about it in here yet:
23:11 < hprmbridge> Katylase> meet Altair! my cubesat friend!^^ https://cdn.discordapp.com/attachments/1064664282450628710/1344567480881774592/IMG_3065.jpg?ex=67c1618c&is=67c0100c&hm=06001e037fa6190d1d2e6e5efcba970a60df4cd7d5741c0c1a27cf01fa91b6fb&
23:11 < fenn> i think every computer science paper dealing with data movement should have a factorio abstract where the algorithm being described is presented as a factorio level, so that the audience can get a visceral concrete feel for what's going on, free of moon runes and vague abstractions
23:12 < fenn> katy i saw altair before, thanks. looks like he's still pretty bare bones
23:12 < fenn> who's the companion cube at the top of the image?
23:14 < fenn> man, i spent so much time on this image but it really works https://fennetic.net/sd/companion_cube2.jpg
23:14 < hprmbridge> Katylase> he's still a prototype (I had to take two years break so I could complete high school... so I didn't have the time for him🥺 )
23:15 < fenn> just make sure you see school as a choice
23:15 < fenn> don't do things just because people tell you to
23:17 < fenn> you always choose whether to trust someone's judgement about what's responsible
23:18 < fenn> honestly i don't know how people are planning for the future these days, it's not like educational institutions are relevant anymore in the most important fields
23:18 < fenn> still good for high capital intensity learning like biotech, robotics, physics
23:19 < fenn> but with stuff like electrical engineering or mechanical engineering you can do 90% of it at home at your own pace
23:20 < fenn> it's just hard to maintain initiative and be honest about your level of effort when there's no external accountability for (lack of) progress
23:21 < fenn> there's surely cheaper ways to pay for external accountability tho
23:21 < fenn> beeminder would donate toward cause you liked (if you accomplished your self-imposed goal) or toward a cause you hated (if you failed at the goal)
23:26 < hprmbridge> Katylase> oh! that's his carry case... like these... https://cdn.discordapp.com/attachments/1064664282450628710/1344571251900940308/IMG_3066.jpg?ex=67c1650f&is=67c0138f&hm=9f75d7b3ca265095c7d7923a1161314aa98a48af51c59a2b18a42bce244d1a2d&
23:27 < hprmbridge> Katylase> i actually https://cdn.discordapp.com/attachments/1064664282450628710/1344571482680066088/IMG_3067.jpg?ex=67c16546&is=67c013c6&hm=7b67efc03cc033bfe33342465de7e11e5b12be3b2eb0c3b38f3e31a7645ae781&
23:27 < hprmbridge> Katylase>  https://cdn.discordapp.com/attachments/1064664282450628710/1344571715153432596/IMG_3067.jpg?ex=67c1657d&is=67c013fd&hm=b90e12ae19128da8774a07c4226943a6ec1206c95fcb6e372ff3852f36f457fe&
23:29 < hprmbridge> Katylase> i actually liked high school, it's just the end exams were too stressful...
23:39 < jrayhawk> Colleges/Universities are also concentrations of intellectuals/experts/students within a given field
23:39 < fenn> yes school is nice in that everyone is there on roughly the same terms so you get an instant "in" to the nascent social network, unlike the real world where you usually have to work to get into a scene
23:40 < jrayhawk> which, notably, you don't typically even need to pay to get access to
23:40 < fenn> been hearing "you can just do things" a lot lately
23:41 < fenn> you can just write emails to professors, usually they will even read it
23:41 < fenn> have ChatGPT critique the email to sound like less of a dumbass
23:42 < fenn> the price of peace is at an all time low!
23:43 < jrayhawk> student groups, meeting spaces, and classrooms typically have no mandatory access control
23:43 < jrayhawk> you can just waltz right in
23:44 < jrayhawk> lab spaces usually do, though, but you just have to make friends for those
--- Log closed Thu Feb 27 00:00:30 2025