--- Log opened Tue Oct 07 00:00:01 2025
00:20 -!- Jenda_ [~jenda@coralmyn.hrach.eu] has joined #hplusroadmap
00:20 -!- Jenda [~jenda@coralmyn.hrach.eu] has quit [Ping timeout: 256 seconds]
00:52 -!- stipa [~stipa@user/stipa] has quit [Read error: Connection reset by peer]
00:52 -!- stipa [~stipa@user/stipa] has joined #hplusroadmap
01:21 -!- Croran_ [~Croran@user/Croran] has quit [Ping timeout: 264 seconds]
01:24 -!- gl00ten [~gl00ten@2001:8a0:7ee5:7800:46d9:f5c:17a2:432] has joined #hplusroadmap
01:37 -!- Croran [~Croran@user/Croran] has joined #hplusroadmap
02:57 -!- hellleshin [~talinck@172-2-128-195.lightspeed.cntmoh.sbcglobal.net] has quit [Ping timeout: 256 seconds]
04:01 -!- srk [~sorki@user/srk] has quit [Ping timeout: 264 seconds]
04:02 -!- srk [~sorki@user/srk] has joined #hplusroadmap
04:15 -!- redlegion_ is now known as redlegion
04:33 -!- TMM_ [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
04:33 -!- TMM [hp@amanda.tmm.cx] has joined #hplusroadmap
04:58 < hprmbridge> kanzure> someone should run a mouse longevity escape velocity project. to my knowledge everyone does separate experiments on mice because otherwise you can't reason about causation or effect sizes. but then you don't try the LEV thing!
04:59 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has quit [Quit: Avoid fossil fuels and animal products. Have no/fewer children. Protest, elect sane politicians. Invest ecologically.]
05:00 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has joined #hplusroadmap
05:03 < hprmbridge> kanzure> "A specialized language model for document information extraction" https://arxiv.org/abs/2509.22906 (7b)
05:06 < hprmbridge> kanzure> some chinese open source LLMs: DeepSeek V3.2, GLM-4.6 (335B-A32B), Qwen3-VL-30B-A3B (Instruct & Thinking), Qwen3-VL-235B-A22B (Instruct & Thinking), Qwen3-Next 80B-A3B (Instruct & Thinking), GLM-4.5V (VLM, 106B-A12B), DeepSeek V3.1, Doubao 1.6-Vision (multimodal, tool-calling), Doubao Translation 1.5 (ByteDance, 28 Languages), ERNIE X1.1 (Baidu, Reasoning), Hunyuan-MT-7B & Chimera-7B (Tencent
05:06 < hprmbridge> kanzure> Translation Specialists), MiniCPM-V 4.5 (8B), Tiny but GPT-4o-level VLM, InternVL 3.5 (MASSIVE Multimodal Family of Models, 1B to 241B Sizes), Step-3 (VLM, 321B/38B), SenseNova V6.5 (SenseTime, Multimodal), GLM-4.5 Air (Base & Instruct, 106B-A12B), GLM-4.5 (Base & Instruct, 335B-A32B), Qwen 3-Coder-30B-A3B (Instruct & Thinking), Qwen3-Coder-480B-A35B (Instruct & Thinking), Qwen3-30B-A3B-2507
05:06 < hprmbridge> kanzure> (Instruct & Thinking), Qwen3-235B-A22B-2507 (Instruct & Thinking), Kimi K2 (1T-A32B)
05:09 < hprmbridge> kanzure> where is longcat in that list? or Mistral?
05:09 -!- helleshin [~talinck@172-2-128-195.lightspeed.cntmoh.sbcglobal.net] has joined #hplusroadmap
05:14 -!- helleshin [~talinck@172-2-128-195.lightspeed.cntmoh.sbcglobal.net] has quit [Read error: Connection reset by peer]
05:18 < hprmbridge> Eli> They have an award for getting mice to live 5 years but no one has accomplished it yet. Very difficult. Methuselah Mouse Prize
05:19 < hprmbridge> Eli> I hear what you’re saying. I think for a portion of the population they can work even if the data is messy.
05:21 < hprmbridge> kanzure> methuselah prize will not itself create the situation where someone does mouse LEV... probably most people think about it as a single intervention bc you will only be able to do multiple interventions outside academia, which excludes almost everyone who works on mice or longevity.
05:24 < hprmbridge> kanzure> given the large investment in mouse science and the difficulty of clinical translation due to either science mismatch or regulatory friction, we should instead migrate the human species from primate biology to mouse biology so that we can better take advantage of our overinvestment.
05:24 < hprmbridge> kanzure> i know i previously said pig is the future, but realistically mouse is a far more studied platform.
05:43 -!- TMM [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
05:43 -!- TMM [hp@amanda.tmm.cx] has joined #hplusroadmap
06:04 < fenn> capybaras-r-us
06:10 < fenn> i don't see a necessary disconnect between academia and multiple running interventions. there are some really long term studies
06:11 < hprmbridge> kanzure> have they done multiple interventions for anything in mice ever?
06:12 < fenn> .t https://orionsarm.com/eg-article/4ba1012793821
06:12 < saxo> Orion's Arm - Encyclopedia Galactica - PackRat Spores, AKA  '____' Spores
06:19 < fenn> the bushbot tail is from rocheworld by robert forward and the book had a nice illustration but you'll have to be satisfied with text http://technovelgy.com/ct/content.asp?Bnum=1777
06:19 < fenn> later popularized by moravec
06:20 -!- helleshin [~talinck@172-2-128-195.lightspeed.cntmoh.sbcglobal.net] has joined #hplusroadmap
06:24 < hprmbridge> kanzure> surely we are living in a golden age of visualizations for orion's arm.
06:26 < hprmbridge> kanzure> pasky, can you add that for this channel
06:27 < fenn> https://fennetic.net/irc/bushbot_rocheworld.png
06:30 < fenn> it's very hard to get the kind of concept-accurate technical detail needed for useful sci-fi illustrations out of a bare image model. you have to do 3d modeling or sketching first, or maybe a very wide nested agent loop of designing, musing, controlnet, composition of multiple images with controlnet, looking at the image to spot errors, until maybe you're done and can show the user
06:31 < fenn> it's getting easier with image editing models but still not good enough to just poop out a fully worked out concept
06:32 < fenn> really cross attention would be best, there was some image model focused on that recently released
06:37 < pasky> kanzure: you mean image generation?
06:41 < hprmbridge> kanzure> an artist that sketches what he sees
06:45 -!- gl00ten [~gl00ten@2001:8a0:7ee5:7800:46d9:f5c:17a2:432] has quit [Ping timeout: 246 seconds]
06:52 < pasky> ...so, image generation? :) i could concievably add something like nanobanana as a tool
06:53 < hprmbridge> kanzure> how about unprompted (make the agent decide if something being discussed is something that it wants to visualize)?
06:57 < fenn> wow i am not a fan of the new qwen blog and the halfassed mandatory meta refresh on the old blog that breaks all the old posts
06:58 < fenn> qwen-image-edit-2509 was the composition model i was thinking of https://qwen.ai/blog?id=1675c295dc29dd31073e5b3f72876e9d684e41c6&from=research.research-list
06:59 < fenn> so you'd design simple elements first on their own, then compose them hierarchically to make more complex scenes without stumbling all over yourself in the prompt
07:00 < fenn> "isn't this just photoshop?" no it's not, because you can re-pose or re-orient people and objects and the lighting matches
07:03 < kanzure> hmm i should throw an agent at the genetic-modifications.mdwn file sometime. it would be interesting to see what several dollars of compute would come up with.
07:16 < hprmbridge> Eli> Yeah, the ITP is doing stuff. They recently used Rapamycin + Trametinib to get almost 30% lifespan increase.
07:16 < hprmbridge> Eli>
07:16 < hprmbridge> Eli> My current hypothesis is that aging is a complex system that has multiple dynamic variables. It's like modeling the economy. You have big variables like unemployment and the federal interest rate, then you also have external shocks and a bunch of smaller variables and they all interact with each other. I'm guessing same thing with the body. You have to impact a lot of parts of the body
07:16 < hprmbridge> Eli> simultaneously, and some of the variables are probably of significantly greater importance. But also, I'm an idiot, so who knows? Probably embryonic gene editing get's you highest ROI. If you make it past 100, it's probably due to genetic reasons.
07:16 < hprmbridge> Eli>
07:16 < hprmbridge> Eli> Mice are also very different than humans. They are a prey species and evolution made them basically speedrun growth and babies. We should hypothesize that treatments in mice won't exactly translate 1 to 1 for humans in terms of effect size. For the same reason that modulating IGF1 in small dogs might not be expected to have the same lifespan extending effects that modulating IGF1 in big dogs would
07:16 < hprmbridge> Eli> have.
07:18 < kanzure> ITP?
07:19 < fenn> "due to genetic reasons" is just punting the mechanism elucidation down the road
07:19 < fenn> *everything* is due to genetic reasons
07:19 < kanzure> i'm a big fan of monocausal reasoning. reject multicausal epistemology before exponential blowup kills us all!!
07:20 < fenn> reality is disgustingly complex and if you want to pretend otherwise you're one of Them
07:20 < fenn> see things as they are, not as you wish them to be
07:21 < kanzure> so you admit your mechanism elucidation might actually be multicausal mechanisms
07:21 < fenn> of course
07:21 -!- gl00ten [~gl00ten@194.117.18.99] has joined #hplusroadmap
07:22 < fenn> we should strive to eliminate monocausal hypotheses first, because they have the most explanatory power and fewest solomonoff induction bits
07:22 < fenn> also, i am not awake
07:22 < fenn> what are you people doing in my bed
07:24 -!- gl00ten2 [~gl00ten@nat-vpn-fm.net.ul.pt] has joined #hplusroadmap
07:27 -!- gl00ten [~gl00ten@194.117.18.99] has quit [Ping timeout: 256 seconds]
07:50 < pasky> │ Hi, let's extend the current set of LLM tools with a nanobanana tool - invoking https://openrouter.ai/google/gemini-2.5-flash-preview-image model with the current chat history to      │
07:50 < pasky> │ generate an image, then storing it as an artifact. Discuss design & implementation & review with the oracle.                                                                            │
07:50 < pasky> let's see what happens :)
07:58 < hprmbridge> Eli> Interventions Testing Protocol. They test life extending drugs on non-inbred mice across three labs. https://www.nia.nih.gov/research/dab/interventions-testing-program-itp/supported-interventions
07:59 -!- gl00ten3 [~gl00ten@nat-vpn-fm.net.ul.pt] has joined #hplusroadmap
08:01 -!- gl00ten2 [~gl00ten@nat-vpn-fm.net.ul.pt] has quit [Ping timeout: 256 seconds]
08:02 -!- gl00ten [~gl00ten@nat-vpn-fm.net.ul.pt] has joined #hplusroadmap
08:04 -!- gl00ten3 [~gl00ten@nat-vpn-fm.net.ul.pt] has quit [Ping timeout: 264 seconds]
08:31 < kanzure> surgical plan: bilateral anterior capsulotomy, targeting the ventromedial anterior limb of the internal capsule to modulate orbitofronto-thalamic and cingulo-striatal pathways while preserving dorsolateral associative and posterior fibers. Surgeon should review tractography and microanatomy of: 1) Frontothalamic fibers connecting the ventromedial prefrontal cortex (BA 11, 12, 25) and anterior ...
08:31 < kanzure> ...cingulate (BA 24, 32) to the mediodorsal and anterior thalamic nuclei. 2) Ventral striatal projections coursing anterior-inferior in the capsule (near the nucleus accumbens and caudate head). 3) The dorsal boundary of the target region relative to caudate nucleus and putamen, and the posterior limit near the genu of the internal capsule, to avoid involvement of frontopontine and motor fibers. ...
08:31 < kanzure> ...4) Stereotactic landmarks: typically 7–10 mm anterior to the anterior commissure, 2–3 mm lateral to the third ventricle wall, 45–50° from the midline plane (verify per patient MRI/DTI).
08:31 < kanzure> patients undergoing this procedure demonstrate a 5 to 10 IQ point increase postoperatively. however there might be severe selection effects where these patients have an actual pathology and wouldn't otherwise experience this increase without that pathology under the same neurosurgery.
08:32 < kanzure> wow apparently there is no "QALY per IQ point" estimates in the literature? what?
08:33 < jrayhawk> Not even derived from observational data?
08:33 < kanzure> about 20% lower all-cause mortality hazard per standard deviation (15 IQ points)... at least that's something. my QALY probably goes up if i'm actually alive?
08:38 < jrayhawk> i dunno, does "being slowly devoured by fleas in the lubyanka because the intelligentsia can't be trusted to have the interests of the common people" count as negative quality?
08:40 -!- gl00ten [~gl00ten@nat-vpn-fm.net.ul.pt] has quit [Ping timeout: 264 seconds]
08:46 < kanzure> oh, focused ultrasound capsulotomy clinical trials have happened. cool.
08:47 < kanzure> https://www.sciencedirect.com/science/article/abs/pii/S0006322324015488
09:10 -!- gl00ten [~gl00ten@nat-vpn-fm.net.ul.pt] has joined #hplusroadmap
09:32 -!- gl00ten [~gl00ten@nat-vpn-fm.net.ul.pt] has quit [Ping timeout: 246 seconds]
09:36 -!- gl00ten [~gl00ten@nat-vpn-fm.net.ul.pt] has joined #hplusroadmap
10:10 -!- gl00ten [~gl00ten@nat-vpn-fm.net.ul.pt] has quit [Ping timeout: 246 seconds]
10:13 -!- gl00ten [~gl00ten@194.117.18.99] has joined #hplusroadmap
10:29 -!- hellleshin [~talinck@172-2-128-195.lightspeed.cntmoh.sbcglobal.net] has joined #hplusroadmap
10:29 -!- stipa_ [~stipa@user/stipa] has joined #hplusroadmap
10:29 -!- stipa [~stipa@user/stipa] has quit [Read error: Connection reset by peer]
10:29 -!- gl00ten [~gl00ten@194.117.18.99] has quit [Read error: Connection reset by peer]
10:29 -!- stipa_ is now known as stipa
10:30 -!- gl00ten [~gl00ten@194.117.18.99] has joined #hplusroadmap
10:32 -!- helleshin [~talinck@172-2-128-195.lightspeed.cntmoh.sbcglobal.net] has quit [Ping timeout: 256 seconds]
10:43 -!- A_Dragon [A_D@libera/staff/dragon] has joined #hplusroadmap
10:44 -!- redlegion_ [~x@omghax.redlegion.org] has joined #hplusroadmap
10:46 -!- MuaddibL1M [muaddib@pasky.or.cz] has joined #hplusroadmap
10:51 -!- Netsplit *.net <-> *.split quits: gAy_Dragon, redlegion, MuaddibLLM
11:04 -!- A_Dragon [A_D@libera/staff/dragon] has quit [Quit: ZNC - https://znc.in]
11:04 -!- A_Dragon [A_D@libera/staff/dragon] has joined #hplusroadmap
11:04 -!- A_Dragon is now known as Awoobis
11:12 -!- redlegion_ is now known as redlegion
11:30 -!- MuaddibL1M is now known as MuaddibLLM
11:32 < pasky> MuaddibLLM: illustate the surgical plan
11:33 < MuaddibLLM> cranking out a nice capsulotomy schematic for ya...
11:34 < MuaddibLLM> got the image, now packaging it up as an artifact...
11:34 < MuaddibLLM> Here's your capsulotomy schematic - image at https://pasky.or.cz/muaddib/242f267963384602b4272b9b455f5ba2.png with full annotations at https://pasky.or.cz/muaddib/644f50dd64524b5ea31a023c45056271.txt
12:00 < kanzure> pasky: can i request an imagemagick step that adds a "slop" symbol on these images somewhere? thx
12:01 < kanzure> these links will be auto-archived by fenn's gallery tool (made because discord image links are too ephemeral)
12:03 < pasky> hehe it looks too real at the first sight, right?
12:05 < kanzure> pasky: what has your hplusroadmap cost burn (in tokens or API call costs) been so far?
12:06 < pasky> i don't really account per channel
12:06 < kanzure> hear that boys? fire away!
12:06 < pasky> but overall Muaddib consumption is roughly $50/month right now
12:08 -!- helleshin [~talinck@172-2-128-195.lightspeed.cntmoh.sbcglobal.net] has joined #hplusroadmap
12:09 -!- hellleshin [~talinck@172-2-128-195.lightspeed.cntmoh.sbcglobal.net] has quit [Ping timeout: 256 seconds]
12:11 -!- hellleshin [~talinck@172-2-128-195.lightspeed.cntmoh.sbcglobal.net] has joined #hplusroadmap
12:13 -!- helleshin [~talinck@172-2-128-195.lightspeed.cntmoh.sbcglobal.net] has quit [Ping timeout: 240 seconds]
13:56 -!- flooded [~flooded@149.102.226.202] has quit [Remote host closed the connection]
13:57 -!- flooded [~flooded@149.102.226.202] has joined #hplusroadmap
14:22 -!- TC [~talinck@172-2-128-195.lightspeed.cntmoh.sbcglobal.net] has joined #hplusroadmap
14:23 -!- gl00ten2 [~gl00ten@194.117.18.99] has joined #hplusroadmap
14:25 -!- gl00ten [~gl00ten@194.117.18.99] has quit [Read error: Connection reset by peer]
14:25 -!- TMM [hp@amanda.tmm.cx] has quit [Ping timeout: 256 seconds]
14:25 -!- srk [~sorki@user/srk] has quit [Ping timeout: 256 seconds]
14:25 -!- TMM_ [hp@amanda.tmm.cx] has joined #hplusroadmap
14:25 -!- srk [~sorki@user/srk] has joined #hplusroadmap
14:25 -!- hellleshin [~talinck@172-2-128-195.lightspeed.cntmoh.sbcglobal.net] has quit [Ping timeout: 256 seconds]
14:28 -!- flooded [~flooded@149.102.226.202] has quit [Remote host closed the connection]
14:31 -!- flooded [~flooded@149.102.226.202] has joined #hplusroadmap
14:39 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has quit [Ping timeout: 264 seconds]
16:41 -!- Gooberpatrol66 [~Gooberpat@user/gooberpatrol66] has quit [Quit: Konversation terminated!]
16:50 -!- Gooberpatrol66 [~Gooberpat@user/gooberpatrol66] has joined #hplusroadmap
16:56 -!- stipa_ [~stipa@user/stipa] has joined #hplusroadmap
16:58 -!- stipa [~stipa@user/stipa] has quit [Ping timeout: 256 seconds]
16:58 -!- stipa_ is now known as stipa
17:35 -!- flooded [~flooded@149.102.226.202] has quit [Remote host closed the connection]
17:36 -!- flooded [~flooded@149.102.226.202] has joined #hplusroadmap
17:37 -!- abetusk [~abetusk@syn-024-059-059-088.res.spectrum.com] has joined #hplusroadmap
18:37 -!- flooded [~flooded@149.102.226.202] has quit [Remote host closed the connection]
18:38 -!- flooded [~flooded@149.102.226.202] has joined #hplusroadmap
18:46 -!- gl00ten2 [~gl00ten@194.117.18.99] has quit [Ping timeout: 264 seconds]
18:55 -!- abetusk_ [~abetusk@syn-024-059-059-088.res.spectrum.com] has joined #hplusroadmap
18:57 -!- abetusk [~abetusk@syn-024-059-059-088.res.spectrum.com] has quit [Read error: Connection reset by peer]
18:58 -!- MuaddibLLM [muaddib@pasky.or.cz] has quit [Ping timeout: 256 seconds]
18:58 -!- MuaddibLLM [muaddib@pasky.or.cz] has joined #hplusroadmap
19:29 -!- TMM_ [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
19:29 -!- TMM [hp@amanda.tmm.cx] has joined #hplusroadmap
19:34 -!- flooded [~flooded@149.102.226.202] has quit [Remote host closed the connection]
19:40 -!- flooded [~flooded@149.102.226.202] has joined #hplusroadmap
19:41 -!- flooded [~flooded@149.102.226.202] has quit [Remote host closed the connection]
19:52 -!- abetusk_ is now known as abetusk
20:21 < hprmbridge> kanzure> AlonzoTg isn't dead? https://agi.topicbox.com/groups/agi/T42f89dc6f956dc87/taking-a-shot
20:23 < hprmbridge> kanzure>  https://cdn.discordapp.com/attachments/1064664282450628710/1425322603538546748/image0.png?ex=68e72a9d&is=68e5d91d&hm=9b270b0ab53cc3659f5cd5c572154c875e0c8894cd3c6b3150563f325a7d3b28&
20:28 < hprmbridge> kanzure>  https://cdn.discordapp.com/attachments/1064664282450628710/1425323885431099463/image0.png?ex=68e72bcf&is=68e5da4f&hm=e08eaafbead0316065c95f1acdc481f2e0ef529ab78cf3789c96f9ce67025a3e&
20:28 < hprmbridge> kanzure>  https://cdn.discordapp.com/attachments/1064664282450628710/1425323992121610330/image0.png?ex=68e72be8&is=68e5da68&hm=657e832f3db9dd0677c1c40f0d9631af831fe0ae9e004c50e7cfd3f030049b5c&
20:36 < hprmbridge> kanzure> https://bioportal.bioontology.org/ontologies/FMA/?p=classes&conceptid=http%3A%2F%2Fpurl.org%2Fsig%2Font%2Ffma%2Ffma83848
20:47 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has quit [Ping timeout: 248 seconds]
22:04 < fenn> really i should be auto-scraping *every* url that comes through here
22:41 -!- redlegion [~x@omghax.redlegion.org] has quit [Remote host closed the connection]
22:41 -!- redlegion [~x@omghax.redlegion.org] has joined #hplusroadmap
23:30 < fenn> ok i think it's scraping muaddib now
23:33  * fenn tries to read "dis-dis-inhibitory" and fails
23:33 < fenn> pathetic
23:34 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has joined #hplusroadmap
--- Log closed Wed Oct 08 00:00:02 2025