--- Log opened Tue Oct 28 00:00:21 2025
00:53 < hprmbridge> kanzure> "Tensor product attention is all you need" https://arxiv.org/abs/2501.06425
01:21 < hprmbridge> kanzure> grokipedia launched  https://grokipedia.com/page/Race_and_intelligence
01:22 < fenn> soft exoskeleton prototype https://andrea-locatelli.com/seismic-clothing
01:23 < fenn> "Articles Available 885,279" jeeez
01:26 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has joined #hplusroadmap
01:37 -!- TMM [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
01:37 -!- TMM [hp@amanda.tmm.cx] has joined #hplusroadmap
02:50 < hprmbridge> kanzure> https://old.reddit.com/r/LocalLLaMA/comments/1o5ptit/ring1t_the_opensource_trillionparameter_thinking/
02:51 < hprmbridge> kanzure> https://github.com/Aider-AI/aider/issues/4584
02:56 < hprmbridge> kanzure> what's a good alternative? qwen-code?
03:00 < hprmbridge> kanzure> ampcode is not open source?
03:08 < hprmbridge> kanzure> https://github.com/QwenLM/Qwen3-Coder
03:09 < hprmbridge> kanzure> oops i meant https://github.com/QwenLM/qwen-code
03:12 < hprmbridge> kanzure> https://ampcode.com/
03:21 < fenn> "Qwen Code is ... adapted from Gemini CLI" heard many bad things about "Gemini CLI"
05:34 < hprmbridge> kanzure> "Resolution limit of the eye: how many pixels can we see?" https://www.nature.com/articles/s41467-025-64679-2 "We measure achromatic (black-white) and chromatic (red-green and yellow-violet) resolution limits for foveal vision, and at two eccentricities (10° and 20°). Our results demonstrate that the resolution limit is higher than what was previously believed, reaching 94 pixels per degree (ppd)
05:34 < hprmbridge> kanzure> for foveal achromatic vision, 89 ppd for red-green patterns, and 53 ppd for yellow-violet patterns. We also observe a much larger drop in the resolution limit for chromatic patterns (red-green and yellow-violet) than for achromatic patterns."
05:34 < hprmbridge> kanzure> let's add more axons into the optic nerve. and more retinal ganglion cells.
05:48 < hprmbridge> kanzure> dinosaur feathers preserved in amber, not enough for an outfit sadly https://cdn.discordapp.com/attachments/1064664282450628710/1432712696192831528/image0.jpg?ex=69020d2d&is=6900bbad&hm=e76c1db02cdc3211349e76890f11099bc8b1556b034f10ad722d2b1e70f8e0b7&
05:58 < hprmbridge> kanzure> "Niron's rare-earth-free permanent magnets use iron-nitrogen chemistry to form iron nitride nanoparticles (Fe16N2) (or very fine powders) with controlled size, orientation and structure (to maximise alignment, anisotropy, magnetic coupling). These powders are then pressed, compacted, possibly sintered (or otherwise consolidated) into bulk magnet shapes (rods, blocks, whatever the motor or device
05:58 < hprmbridge> kanzure> demands) with alignment of magnetic grains. Coercivity (ability to resist demagnetisation) may be lower than the best rare-earth magnets. The iron nitride material has high saturation magnetization: the iron atoms, with nitrogen interstitials, lead to stronger magnetic moment per volume than many alternative permanent magnet materials."
06:13 -!- etc-vi [~etc-vi@user/meow/girlchunks] has quit [Ping timeout: 256 seconds]
06:20 -!- etc-vi [~etc-vi@user/meow/girlchunks] has joined #hplusroadmap
09:16 < kanzure> "Natural induction: Spontaneous adaptive organization without natural selection" https://www.mdpi.com/1099-4300/26/9/765 (2024)
09:34 -!- 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.]
09:35 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has joined #hplusroadmap
09:45 < kanzure> lesswrong posts are now showing up in google scholar search results. probably a metatag trick.
09:45 < hprmbridge> kanzure> "adaptation by natural induction" diagram https://cdn.discordapp.com/attachments/1064664282450628710/1432772377221271692/go9jhkhrcmd5uzvckmcd.png?ex=690244c2&is=6900f342&hm=5919e8455ba78f9462a986ec22e9561fe7c688b990d9ed4351708030e7021c5f&
10:43 < kanzure> "Goal-directed learning in cortical organoids" https://www.biorxiv.org/content/10.1101/2024.12.07.627350v1.abstract (2024) from "Open Culture Science, Inc" -- " We developed a closed-loop electrophysiology framework to embody mouse cortical organoids into a simulated dynamical task (the inverted pendulum problem known as ‘Cartpole’) and evaluate learning through high-frequency training ...
10:43 < kanzure> ...signals. Longitudinal experiments enabled by this framework illuminate how different methods of selecting training signals enable improvement on the tasks. We found that for most organoids, training signals chosen by artificial reinforcement learning yield better performance on the task than randomly chosen training signals or the absence of a training signal."
10:51 < kanzure> "神経細胞の分散培養系の情報処理容量" https://www.jstage.jst.go.jp/article/ieejeiss/142/5/142_578/_article/-char/ja/ suwa et al 2022 showing a neural tissue culture performing arithmetic operations.
10:57 < kanzure> "From neurons to computation: Biological reservoir computing for pattern recognition" https://arxiv.org/abs/2505.03510 (2025) showing hippocampal neuron culture performing "positional codes, bars with different orientations, and a [pixel image] digit recognition task"
10:57 < kanzure> oh, "classifying handwritten digits (N-MNIST) with high accuracy". okay. handwritten then.
11:01 < hprmbridge> kanzure> "synthetic biological intelligence" diagram https://cdn.discordapp.com/attachments/1064664282450628710/1432791478056390687/gr1_lrg.png?ex=6902568c&is=6901050c&hm=457a70541a207719924f0741b011df9ba0dd65e67704d2b424eda8578d105e7c&
11:04 -!- TMM [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
11:04 -!- TMM [hp@amanda.tmm.cx] has joined #hplusroadmap
11:08 < kanzure> "Two roads diverged: Pathways toward harnessing intelligence in neural cell cultures" https://www.cell.com/cell-biomaterials/fulltext/S3050-5623(25)00147-3 (2025)
11:18 < kanzure> bj kagan from cortical labs has coauthored with savulescu. hrm.
11:48 < kanzure> "Supervised learning in DNA neural networks" https://www.nature.com/articles/s41586-025-09479-w (2025) 
11:48 < kanzure> "Here we show that DNA molecules can be programmed to autonomously carry out supervised learning in vitro, with the system learning to perform pattern classification from molecular examples of inputs and desired responses. We demonstrate a DNA neural network trained to classify three different sets of 100-bit patterns, integrating training data directly into memories of molecular ...
11:48 < kanzure> ...concentrations and using these memories to process subsequent test data. Our work suggests that molecular circuits can learn tasks more complex than simple adaptive behaviours."
11:57 < kanzure> "Physics-driven learning network made of self-adjusting variable resistors" https://diyhpl.us/~bryan/papers2/ai/physical-intelligence/Laboratory%20demonstration%20of%20decentralized%20physics-driven%20learning%20-%202021.pdf
12:03 < kanzure> mechanical neural network for handwritten digit recognition task https://advanced.onlinelibrary.wiley.com/doi/full/10.1002/advs.202308137
12:04 < hprmbridge> kanzure>  https://cdn.discordapp.com/attachments/1064664282450628710/1432807147753836624/advs7282-fig-0001-m.png?ex=69026524&is=690113a4&hm=06a00a70c97f946b7895130a0298e9ed0f1b005fc4bed3ef898f49b993f14898&
12:04 < hprmbridge> kanzure>  https://cdn.discordapp.com/attachments/1064664282450628710/1432807198286676048/advs7282-fig-0004-m.png?ex=69026530&is=690113b0&hm=33e2270484d92c35e990fa798d733cdac5cef149197b5a15a641b2cd62a9ac93&
12:07 < kanzure> maybe with SMT pick-and-place you could encode a very small LLM into uh... um. well something that is not electrical in nature would be nice from a non-silicon-dependency angle, but i'm pretty sure most options are going to result in extremely slow computation.
12:08 < kanzure> karpathy was recently posting about $1000 training cost for super tiny LLMs as a toy model to play with. you can even chat with them and see that they have at least some semblance of a personality.
12:12 < kanzure> what is a "kirigami mechanical sensor"?
12:30 < kanzure> "a low-energy-loss solitary wave-based mechanical metamaterial structure optimized for efficient information transfer and logic gate implementation" https://www.nature.com/articles/s41467-024-47201-y
12:31 < hprmbridge> kanzure>  https://cdn.discordapp.com/attachments/1064664282450628710/1432813937216196668/41467_2024_47201_Fig5_HTML.png?ex=69026b77&is=690119f7&hm=9cf65541437ea9ba4f098b945dbf45f414caed4095ee3513643f3a802698d22e&
12:31 < MuaddibLLM> [gpt-5] Kirigami mechanical sensors use strategic cut patterns (inspired by Japanese paper-cutting art) in thin materials to create stretchable, conformable structures for sensing strain, pressure, or deformation - think precise geometric cuts in graphene or piezoelectric sheets that unfold/morph under load https://onlinelibrary.wiley.com/doi/full/10.1002/advs.202204733
12:38 < hprmbridge> kanzure>  https://cdn.discordapp.com/attachments/1064664282450628710/1432815836191653999/image.png?ex=69026d3b&is=69011bbb&hm=e37cc652a026862100bdb4747abd94edc3b12fcc1a12dc0a2a33ef09b29a8956&
13:07 < kanzure> "KirchhoffNet" proposes an analog circuit that generates diffusion images at a rate of 1 image/picosecond https://arxiv.org/abs/2310.15872 https://zhengqigao.github.io/articles/what_is_kirchhoffnet.pdf
13:25 < juri_> that's... a lot of not doing math, or being reproduceable.
14:15 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=7d84038f Bryan Bishop: more on multiplex addressing of cells >> http://diyhpl.us/diyhpluswiki/senses/
14:18 < kanzure> here is someone working on molecular information systems, self-replication, physical learning systems, some winfree/szostak intersection stuff, https://muruganlab.uchicago.edu/
14:19 < kanzure> "Suppression of errors in collectively coded information" https://arxiv.org/abs/2508.21806 using a simplified model of origin of life scenario to show information propagation and maintenance.
14:21 < kanzure> "A minimal scenario for the origin of non-equilibrium order" https://arxiv.org/abs/2405.10911 where murugan and szostak argue that if the distribution of replication times is wide enough then non-equilibrium order can evolve before that order is directly functional, through selection for fast replication alone. experimentally verified with a mutational screen against proofreading polymerases.
15:54 < hprmbridge> kanzure> buy a NEO robot https://www.1x.tech/order
15:55 < L29Ah> does it have FOSS firmware?
16:06 < andytoshi> L29Ah: lmao
16:06 < andytoshi> it has a 1x employee constantly surveilling the inside of your house https://1x.recruitee.com/o/robot-operator
16:47 < hprmbridge> kanzure> https://grokipedia.com/page/Mind_uploading
17:21 < jrayhawk> ugh, the gliadin article is ass-backwards, "streetlight effect"ing epistemic obviousness and epistemic progression into importance/emphasis. I guess I'll see how well this AI edit suggestion thing works.
17:23 < jrayhawk> Etiology and importance winds up being the exact opposite of epistemic obviousness/progression.
17:35 -!- stipa_ [~stipa@user/stipa] has joined #hplusroadmap
17:37 -!- stipa [~stipa@user/stipa] has quit [Ping timeout: 264 seconds]
17:37 -!- stipa_ is now known as stipa
17:38 -!- justanotheruser [~justanoth@gateway/tor-sasl/justanotheruser] has quit [Ping timeout: 272 seconds]
17:40 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has quit [Ping timeout: 256 seconds]
17:40 < jrayhawk> At least, in part.
17:42 < jrayhawk> Obviously that doesn't work out as a general principle because it creates conspiratorial brainworms ("What I don't observe intrinsically matters more than what I do observe")
17:44 -!- justanotheruser [~justanoth@gateway/tor-sasl/justanotheruser] has joined #hplusroadmap
17:47 < jrayhawk> Which is, in turn, also not generalizable, as seen with e.g. publication bias and other forms of burying results.
18:15 -!- TMM [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
18:15 -!- TMM [hp@amanda.tmm.cx] has joined #hplusroadmap
18:27 < hprmbridge> kanzure> https://tee.fail/
19:45 -!- _flooded [~flooded@149.88.18.226] has joined #hplusroadmap
19:48 -!- _flood [~flooded@149.88.18.226] has quit [Ping timeout: 256 seconds]
19:55 -!- stipa_ [~stipa@user/stipa] has joined #hplusroadmap
20:02 -!- Netsplit *.net <-> *.split quits: L29Ah, stipa
20:02 -!- stipa_ is now known as stipa
--- Log closed Wed Oct 29 00:00:22 2025