--- Log opened Tue Dec 26 00:00:10 2023
00:57 < fenn> is there some other path back to portable software?
00:59 < fenn> the pytorch dev there is complaining about being constrained to "least common denominator" OpenCL, instead of all the vendor lock-in special purpose tools
01:02 < hprmbridge> fodagut> Vulkan, WebGPU
01:04 < fenn> "The primary target devices are mobile GPUs on Android devices." (pytorch vulkan backend)
01:05 < fenn> perhaps i should have been more clear
01:06 < fenn> is there another path back to running academic-grade bleeding edge deep learning code, without being constrained to use the same GPU as the developer
01:07 < fenn> the window of GPU model years supported by the most recent manufacturer provided GPGPU libraries (CUDA, ROCm) is shrinking
01:07 < fenn> when devs write code relying on recent driver functionality, it severely limits the set of GPUs that the code can run on
01:08 < fenn> i didn't think vulkan was specific to android, or even primarily focusing on that segment, so i'm sorta confused
01:09 -!- darsie [~darsie@84-113-55-200.cable.dynamic.surfer.at] has joined #hplusroadmap
01:09 < fenn> maybe android stuff will only run on vulkan, but it's not as good as the other backends. it's used for android stuff as a last resort since there's no other option? but then why are 5 year old video cards abandoned entirely?
01:10 < fenn> they should be supported via vulkan too, no?
01:40 -!- yorick [~yorick@user/yorick] has quit [Remote host closed the connection]
01:41 -!- yorick [~yorick@user/yorick] has joined #hplusroadmap
02:24 -!- srk [~sorki@user/srk] has quit [Quit: ZNC 1.8.1 - https://znc.in]
02:33 -!- srk [~sorki@user/srk] has joined #hplusroadmap
03:13 -!- Ashstar [~Ashstar@mobile-166-137-177-088.mycingular.net] has quit [Read error: Connection reset by peer]
03:14 -!- Ashstar [~Ashstar@mobile-166-137-177-088.mycingular.net] has joined #hplusroadmap
03:52 -!- test_ [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
03:56 -!- flooded [flooded@gateway/vpn/protonvpn/flood/x-43489060] has quit [Ping timeout: 256 seconds]
05:26 -!- yashgaroth [~ffffffff@2605:a601:a0cd:9500:78a0:7f36:588:173] has joined #hplusroadmap
05:57 < hprmbridge> kanzure> deploy the bees https://youtube.com/watch?v=_T7d0bze4kM
05:59 < hprmbridge> kanzure> there is a machine that auto trains bees at scale, can't find its name
06:55 < TMA> fenn: the incentives are against portable software
06:58 < TMA> fenn: IP laws give the manufacturers monopoly power on driver software, which they use to drive further sales of new product
06:59 < muurkha> fenn: conceivably all the best interventions together could extend your life by 67% if we knew what they were, from 72 years to 120, but a 67% extension is still far from posthumanity
07:01 < muurkha> nobody from the 19th century is alive today despite people trying lots of fasting, exercise, healthy diets, etc.
07:01 < muurkha> so this is bullshit:
07:02 < muurkha> 04:22 < Ashstar> our bodies are amazing
07:02 < muurkha> 04:22 < Ashstar> we have a huge capacity to reset our homeostasis
07:02 < muurkha> 04:23 < Ashstar> we just jam it up
07:03 < muurkha> 01900 is only 122 years ago.  recorded history is 5000 years long, that's 2.5% of recorded history, literally everybody from 97.5% of recorded history is already dead
07:04 < TMA> the number of dead people is way less than 97 % though
07:04 < muurkha> true
07:04 < muurkha> but we're talking about macrobiosis here
07:05 < muurkha> agriculture is 12000 years old, so that's 1% of the history of agriculture.  behavioral modernity is at least 40'000 years old, so 0.3% of behavioral modernity
07:48 < hprmbridge> Eli> I’m currently reading a book that is making the argument that aging is an evolutionary adaptation. Not sure I completely buy all of it but it makes some interesting arguments I cant refute. For example, immortal jellyfish don’t appear to age. And certain trees and a type of turtle are less likely to die each year the older they get. In other words, aging in animals doesn’t seem to be a guarantee.
07:56 < hprmbridge> Eli> The argument is that we must have children due to the red queen hypothesis. But, having too many children could result in extinction in times of hardship. Therefore, elderly act as a buffer by consuming resources during times of plenty, preventing too many progency from being born. And then tend to be the ones that die off during times of hardship, allowing progeny to survive. I'm not convinced
07:56 < hprmbridge> Eli> that Mutation Accumulation and Antagonistic Pleitropy can't play a role, but he dismisses them. I would like to read into the papers he claims dismisses those arguments to see if they hold up.
07:58 < muurkha> aging is clearly an evolutionary adaptation, yes
08:09 < hprmbridge> Eli> what is the teleological reason for it?
08:34 < docl> Well, there's also the tumor surpression theory of aging
08:34 < docl> https://pubmed.ncbi.nlm.nih.gov/34637937/
08:36 < docl> that's my currently favored theory of why evolution neglects to remove aging. seems supported by bowhead whales, which have a novel tumor suppression mechism not present in other cetaceans and a longer lifespan
08:40 < docl> it seems to me that longer lifespan (and reproductive window) is higher fitness overall, it's just that if your species isn't fully immune to cancer it can't be measured as such by evolution. so mechanisms that result in aging but delay cancer tend to be conserved
08:43 < docl> the grandmother hypothesis is interesting as well, as fitness is plausibly conserved in the form of childcare provided outside the reproductive window, which would be a stronger effect in species with communicable skills. hence human post-menopausal aging being slow (and evolution doesn't robustly distinguish gender wrt a general mechanism like this, so guys live longer too) 
09:14 < juri_> as long as you're alive, you're competing with your offspring for resources. i'd think evolution kills you, so that your species can evolve faster.
09:16 < docl> only true if there's pressure to evolve faster. maybe viral evolution forces to occur this somewhat
09:16 < docl> assuming reproductive window is indefinite, you can just keep having more kids, making your genes more likely to survive than a competitor who ages
09:24 < nsh> do you know anyone that programs by continuously changing the operation of a running program that never halts?
09:25 < nsh> i suppose in a repl like ipython3 you can get quite far this way
09:26 < nsh> but at some point you're probably inevitably going to invent versions
09:26 < nsh> so mortality ~ versioning
09:27 < nsh> being eager to undo the utility of versioning because we identify with a running and not the endeavour of evolving a codebase
09:27 < nsh> might be a choice subject to some consideration 
09:27 < nsh> a running *instance
09:29 < docl> well, you might handle that at the software level instead of people *dying*
09:31 < nsh> handle what?
09:31 < nsh> there not being versions?
09:31 < nsh> last version syndrome as an analogy to main character syndrome :)
09:32 < nsh> why should i be a step along the way to anything? shouldn't all of the other steps have been leading to me
09:32 < nsh> i mean other than by overwhelming statistics
09:33 < nsh> and the non-specialness of any brief period of time in the cosmos
09:33 < nsh> but there's no reason that we can't think of creating a continuity of experience between instances
09:34 < nsh> that's something akin to having the best of both worlds
09:35 < docl> huh? if you have a genetic problem, and fix it, that's a new version. if your fix is bad, you do want to roll that back. versioning is something biology handles by letting people die, it's not a good system.
09:38 < nsh> so why don't we program that way?
09:38 < nsh> how doesn't the same logic apply
09:38 < nsh> should we be still making changes to some code written for a PDP11 in 1970?
09:39 < nsh> as it's running
09:39 < nsh> without it at any point not running and being replaced by a slightly different version of itself that starts running instead?
09:39 < nsh> i mean, give it a go 
09:39 < nsh> for a year. and let me know how it goes :)
09:45 < docl> "Sin, young man, is when you treat people like things. Including yourself. That’s what sin is."
09:46 < docl> how come deathists always think it's so brilliant to scrap human beings like an obsolete piece of silicon hardware or a broken hair dryer
09:47 < docl> sheesh. people do use version control for programming.
09:56 < hprmbridge> Eli> This seems to neglect many organisms like trees, immortal jellyfish, and turtles.
09:57 < hprmbridge> Eli> The grandmother hypothesis also ignores many organisms. Single celled organisms also age. Can they behave as grandmothers?
09:57 < docl> no, the GMH doesn't cause aging, it provides a mechanism to select for less of it
09:58 < docl> tumor suppression mechanisms causing aging is plausible though
09:59 < docl> distinct argument: not having good tumor suppression reduces selective pressure to prevent aging
10:53 < hprmbridge> Eli> The immune system has a major impact on cancer prevention. However, telomerase shortens the lives of stem cells that keep the immune system going. Why does the immune system shut down if we want to prevent cancer. There's no reason telomerase couldn't extend the life of these stem cells.
10:53 < hprmbridge> Eli>
10:53 < hprmbridge> Eli> Furthermore, cellular senescence tends to increase inflammation, which is a risk factor for cancer. While cancers tend to have active telomerase, the shutting down of telomerase in aging cells, which results in inflammation, seems to increase cancer. In other words, it's complicated.
11:03 < docl> complex, yes. the immune system aging is a result of the telomerase not being used on cells more often, which prevents cancers from killing people so much before they reproduce. the immune decline is typically after they reproduce, where selective pressure is weaker.
11:07 < hprmbridge> Eli> Why does this have to be the case? Why can't we have a strong immune system throughout life?
11:07 < docl> if you had a tribe of humans that consistently waited until late 30s to reproduce, their immune systems would evolve to last longer (probably the whole organism)
11:08 < hprmbridge> Eli> Ok, but 80 year men can still reproduce.
11:09 < hprmbridge> Eli> Wouldn't the selfish gene hypothesis have an impact here?
11:09 < docl> yeah if no man could reproduce until he turned 80 this would produce a strong pressure to survive and be fit later
11:10 < hprmbridge> Eli> Certainly. I'm just questioning our assumptions here. If 80 year old men had good immune systems, their genes would propagate more and take over.
11:11 < docl> genes of 80 year olds aren't replicated often enough to have much impact. it's typically lower hanging fruit to just have kids younger
11:14 < docl> picture a dynamic system. there are possible mutations that reduce cancer without increasing aging, possible mutations that increase cancer and slow aging, possible genes that speed aging and reduce or delay cancer. evolution isn't great at distinguishing which is which until it impacts reproductive fitness directly. immortality is only a very weak advantage. and less elaborate mutations are more 
11:14 < docl> probably to begin with
11:14 < docl> *probable
11:18 < hprmbridge> Eli> I guess the best way to know would be somehow to test this hypothesis. In certain scenarios, such as high fertility at old age, we should see longer lifespan in organisms. In other organisms with sort fertility windows, we should see rapid aging.
11:19 < docl> I believe this was tested in fruit flies, will see if I can dig it up
11:27 < docl> https://www.nytimes.com/2005/12/06/science/live-longer-with-evolution-evidence-may-lie-in-fruit-flies.html
11:27 -!- darsie [~darsie@84-113-55-200.cable.dynamic.surfer.at] has quit [Ping timeout: 256 seconds]
11:30 < hprmbridge> Eli> im getting a paywall that my normal tricks cant overcome
11:30 < docl> https://onlinelibrary.wiley.com/doi/full/10.1111/j.1420-9101.2005.00966.x // Evolution of late-life fecundity in Drosophila melanogaster
11:32 < hprmbridge> Eli> I did find this paper which does seem to show some relation between fertility later in life and a flatter mortality curve. However, there are also animals that have a very early bell shaped fertility curve and yet live for a very long time (see the rotifer for example). https://cdn.discordapp.com/attachments/1064664282450628710/1189289623998976111/Diversity_of_ageing_across_the_tree_of_life_--
11:32 < hprmbridge> Eli> _Jones_Owen_R._Scheuerlein_Alexander_Salguero-Gomez_--_Nature_7482_505_pages_169-173_2013_dec_--_10.1038_nature12789_--_3b6cf97bfa2eb51af748ac9422088816_--_Annas_Archive.pdf?ex=659d9f46&is=658b2a46&hm=94bf907ba7a0d3974052bb2e7e1df8f0cdf2eb86aadc7856dff4747aea38e550&
13:58 -!- flooded [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
13:59 < hprmbridge> kanzure> https://github.com/openvmp/partcad
14:02 -!- test_ [flooded@gateway/vpn/protonvpn/flood/x-43489060] has quit [Ping timeout: 255 seconds]
14:05 < fenn> .m https://twitter.com/lcamtuf/status/1739347419845087505#m
14:05 < AugustaAva> ​twitter: <lcamtuf> As 2024 dawns, I feel I have no choice but to deliver a stern warning about reliance on computer algorithms we don't understand, such as fast Fourier transform, Huffman coding, or heapsort
14:14 < hprmbridge> fodagut> fenn: if you want performance, your best bet now is to have a generic interface that maps to CUDA, ROCm, and Apple's recently released framework (the name escapes me).
14:14 < hprmbridge> fodagut> if you want open standards, then you need to repeat that vendor-specific work for Vulkan, or ideally WebGPU.
14:14 < hprmbridge> fodagut> OpenCL as a stanard is simply out of date with modern hardware, much like OpenGL. Trying to get it performant is like trying to force a square peg into a round hole.
14:15 < hprmbridge> fodagut> fft, huffman coding, heapsort? I assume that is trolling, lol
14:16 < L29Ah> it's a pointy joke
14:17 < fenn> i don't know what webgpu is, but it has the word "web" in it which makes me automatically distrustful and suspicious
14:17 < L29Ah> fenn: what do you think of webm?
14:17 < hprmbridge> fodagut> ignore the word web here. i hate that they chose that name. it has abosolutely nothing to do with web, or browsers
14:19 < hprmbridge> fodagut> WebGPU is simply the non-Kronos replacement for Vulkan. It's a cross-platform, cross-language, cross-architecture generic GPU programming interface.
14:19 < hprmbridge> fodagut> Alternatively, you can think of it as a cross-platform version of Apple's Metal library, which it effectively is.
14:20 < L29Ah> huh, llvm can target amdgpu
14:20 < hprmbridge> fodagut> One of its many use cases is the browser, where it is given a javascript API and serves as a replacement for WebGL for rendering in canvas elements. Hence the stupid name.
14:20 < hprmbridge> fodagut> See also web assembly.
14:20 < L29Ah> webassembly is that NIH llvm?
14:21 < hprmbridge> fodagut> yeah, or NIH jvm, take your pick
14:21 < hprmbridge> fodagut> it's somewhere between the two
14:22 < hprmbridge> fodagut> I despise the fact that everything is designed to run in a browser these days, so I spent many years confidently ignoring wasm. I advise others to not repeat my mistake.
14:23 < fenn> does anyone distribute signed .wasm files?
14:25 < fenn> i'm not trying to do a gish gallop, but how do i know my distrust of web* initiatives is misplaced?
14:37 < hprmbridge> fodagut> w3c vs kronoos is just boring industry politics and you shouldn't be invested in one group or another?
14:41 < fenn> that's not really the axis i am concerned about
14:41 < hprmbridge> fodagut> that's my point though
14:43 < hprmbridge> fodagut> basically for $reasons, industry efforts to standardize cross-platform APIs has moved from groups like Kronos to the W3C. Hence the many new cross-platform APIs with "web" in their name.
14:43 < L29Ah> w3c is a joke, it has no real power over technology since browsers ignored xhtml, and these days just codifies chromium behavior as i see it
14:44 < fenn> when they change the standard every 5 years it's hard to keep up
14:44 < L29Ah> (and a few other standards that virtually no one uses)
14:44 < hprmbridge> fodagut> W3C has been drastically remade in the wake of WHATWG
14:45 < fenn> i mean there's a new industry consortium with a new thing that does essentially the same thing, but with a new shiny name and website and vague corpo-speak about what it allegedly does
14:45 < fenn> nobody explains why the new standard is better than the old standard
14:45 < fenn> presumably there is some hardware level operation that the old standard doesn't support, but i don't see why we need to re-do literally everything just ot add a new operation/instruction
14:46 < hprmbridge> fodagut> L29Ah: W3C used to think their role was to define standards. the community and marketplace spoke and put them back in their place.
14:46 < fenn> "the community" must be one i don't belong to
14:46 < hprmbridge> fodagut> now their role is to be a meeting point for industry vendors to exchange ideas and codify standards around what industry is creating
14:46 < hprmbridge> fodagut> fenn: WHATWG, look up its history
14:47 < fenn> "The central organizational membership and control of WHATWG – its "Steering Group" – consists of Apple, Mozilla, Google, and Microsoft"
14:47 < hprmbridge> fodagut> i.e. every major browser manufacturer, no?
14:48 < fenn> mozilla has been failing to represent user interests for decades
14:48 < hprmbridge> fodagut> well that's a different argument to have
14:48 < hprmbridge> fodagut> but nominally mozilla representatives have pushed for community viewpoints here
14:49 < docl> https://www.youtube.com/watch?v=RBJlsrN9Qa0 // 3D printing of graphene aerogel
14:49 < fenn> the whole concept of a "living standard" is such bullshit
14:49 < docl> https://sci-hub.se/10.1002/smll.201503524
14:49 < hprmbridge> fodagut> (and Mozilla representatives to WHATWG or W3C are open source devs mostly, not Mozilla Corporation hacks)
14:50 < fenn> why does everyone hate software that "just works" because you haven't touched it since it was last tested
14:50 < fenn> bit rot shold be illegal lol
14:50 < hprmbridge> fodagut> but anyway this is far afield to the original point. if you want to do GPGPU stuff, OpenCL is dead as a doornail. Vulkan is usable but complex, and in the process of being subplanted by WebGPU.
14:51 < hprmbridge> fodagut> If I was doing anything new, I would use WebGPU. Particularly because there is really good C++ and Rust libraries, and browser support is a nice bonus at times.
14:51 < fenn> the answer is obviously "buy a new NVIDIA card"
14:51 < hprmbridge> nmz787> So WebGPU isn't just for Web? Confusing.
14:52 < hprmbridge> fodagut> neither is web assembly. it's really confusing
14:52 < fenn> there are literally millions of times more users than developers
14:53 < fenn> i'm not going to be rewriting pytorch from scratch
14:53 < fenn> this is insane
14:54 < hprmbridge> fodagut> fenn: what are you trying to do?
14:54 < fenn> the thing that started me down this rabbit hole was "why won't stable diffusion run on an RX580"
14:55  * L29Ah decided to stay out of a rat race and runs matrix multiplications on intel avx, ggerganov <3
14:55 < fenn> that's 100x slower for image generation
14:55 < L29Ah> pytorch is a tool for research anyway
14:55 < fenn> a lot of people are using these "research" models for real stuff
14:56 < L29Ah> and image generators are even more uselessly pathetic than language models
14:56 < fenn> and humans are even more uselessly pathetic than image generators
14:56 < fenn> what's your point
14:56  * L29Ah hides to avoid hijacking the webshit thread
14:56 < fenn> it's a self-hijacking thread
14:58 < hprmbridge> fodagut> fenn: I assume you've tried this? https://jingboyang.github.io/rocm_rx580_pytorch.html
14:59 < hprmbridge> fodagut> unfortunately a lot of the higher-level libraries like ROCm use specific features of a GPU that don't exist on old models. this isn't just planned obsolescence.
15:01 < docl> this graphene oxide stuff might be a lot easier to grow stratospheric platforms with, especially if you can make it from CO2. you print a slurry with water to make an ice mixture, then freeze dry and reduce (can be chemical or thermal, seems like UV exposure would probably work for thin layer at a time process)
15:02 < fenn> i'm taking it on faith that these features exist, but it's always vague dangling pointers
15:03 < hprmbridge> nmz787> It could be as simple as no one who's been developing pytorch was using that rx580, so they didn't define the memory map ranges or something. A few simple magic numbers
15:04 < fenn> amd officially discontinued ROCm support for RX580 and pytorch doesn't want to support anything that AMD won't support
15:05 < fenn> all of the GPU vendors pull this trick
15:06 < fenn> it's like adderall-XR
15:06 < fenn> the old version works just as well but you're not allowed to use it, you must use the new patented thing
15:07 < fenn> at least drug patents last 20 years so everyone doesn't have to reinvent the wheel every 5 years
15:08 < hprmbridge> nmz787> I got the non XR recently to be fair
15:10 < fenn> congrats
15:12 < docl> lisdex generic is out in case anyone missed it
15:13 < L29Ah> do doctors get paid to issue prescriptions for patented amphetamine formulations or something?
15:14 < fenn> yes, in informal bribes
15:15 < fenn> you don't have to buy a john deere tractor, or a seimens PLC
15:16 < fenn> or use google chrome, or microsoft windows
15:16 < L29Ah> so gotta outbribe them
15:17 < fenn> but by the time you finish bending over backwards to get all the non-standard stuff to work, you don't have any time left to do anything productive
15:17 < fenn> siemens*
15:18 < fenn> L29Ah: that is "drug seeking behavior" and will get you blacklisted from receiving any medication at all
15:19 < L29Ah> it can be framed differently
15:19 < L29Ah> for example a premium clinic/doctor that boasts its love for generics
15:46 -!- Gooberpatrol66 [~Gooberpat@user/gooberpatrol66] has quit [Ping timeout: 260 seconds]
16:39 < docl> source on the 160 g/m^3 aerographene process https://onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1002%2Fadma.201204576&file=adma_201204576_sm_suppl.pdf
16:39 < docl> seems pretty different from the 3d printable version of aerographene
16:44 < docl> seems like the 180 g/m^3 "aerographite" using the sacrificial zinc oxide nanostructures is probably easier to make, less steps
16:50 < muurkha> fenn: how does the blacklist thing work?
16:52 < muurkha> Eli: I don't think aging is teleological, just evolutionary.  tumor suppression is one issue, but that's just a specific case of the general bottle-wrack vs. spurge-weed issue
16:52 < muurkha> juri_: "evolution kills you so that your species can evolve faster" is the opposite of how evolution works.  evolution favors genes that prevent evolution, not genes that promote evolution
16:53 < muurkha> if by "evolution" we mean "mutation", as in this case
16:53 < muurkha> 17:24 < nsh> do you know anyone that programs by continuously changing the operation of a running program that never halts?
16:53 < muurkha> yes, every Smalltalk programmer
16:53 < muurkha> Emacs is close to that but not quite there; occasionally Emacs instances do die
16:54 < muurkha> all Smalltalk images, as far as I know, are mutated copies of the original Smalltalk image from the 01970s
16:54 < muurkha> Interlisp works this way too, IIRC
16:55 < muurkha> 17:38 < nsh> so why don't we program that way?
16:55 < nsh> pretty sure there are versions of smalltalk programs and most of the things people do with emacs is revise versions of text, albeit with a bunch of everything else integrated because people get so used to the environment it's more fun to suck other things into that than switch to another
16:55 < muurkha> aside from the fact that we do, there are a lot of engineering considerations in programming that are absent in biology
16:55 < nsh> even if you use emacs to read your emails you're reading an email that exists statically 
16:56 < nsh> although i suppose people do work collaboratively on 'live documents' that are evolving 
16:56 < nsh> but actually under the hood they're successions of static things
16:57 < nsh> the thing closest to what people seem to be striving for as immortal is the environment
16:57 < nsh> the internet for instance
16:57 < nsh> hasn't really been rebooted 
16:57 < nsh> maybe a few times in the very early days
16:57 < nsh> but it's been continuously alive now for most of the time it's been a notion
16:58 < muurkha> Eli: stem cells are high-cancer-risk cells, because cellular senescence is very clearly a tumor-suppression adaptation.  also you have the function of telomerase backwards: it lengthens telomeres, extending the lives of cells, not decreasing it.
16:58 < muurkha> there's presumably a balance where if your immune system cells are too good at avoiding tumor-suppression mechanisms, you get leukemia and die
17:00 < muurkha> WebGPU certainly does have a lot to do with the web and browsers: https://caniuse.com/webgpu
17:00 < hprmbridge> Eli> sorry, I meant that telomerase is bound as senescence occurs. Senescense occurs, then results in inflammation and cancer. I'm not completely understanding why the specific stem cells for immune cells have to stop producing new immune cells. You could very well be correct. But, we are sort of making assumptions based on a priori bias. It would be nice to see these hypotheses tested.
17:01 < muurkha> L29Ah: LLVM doesn't attempt to handle untrusted code, wasm does, and it does a great job of it
17:02 < muurkha> WebGPU *isn't* just for the web, and neither is webassembly
17:04 < muurkha> nsh: there are certainly versions of smalltalk programs, but they are all mutated copies of a common memory image that originated almost half a century ago
17:05 < muurkha> Eli: aha, okay
17:05 < muurkha> I agree that I could be wrong about leukemia being the evolutionary reason for immune decline
17:05 < hprmbridge> Eli> For example, why not have the hypothesis that immune stem cells in particular can't unbind telomerase to prevent inflammation and and rapid immune system decline? Immune function is hugely important to cancer prevention. It seems something that would be good to test if possible. Furthermore, we know that mice with shorter telomeres die sooner. And not necessarily from cancer.
17:09 < hprmbridge> Eli> @muurkha, would you be able to post the bottle-wrack vs. spurge-weed site. I remember it was posted here ealier but I cant find it with search.
17:11 < docl> https://avalonlibrary.net/ebooks/Richard%20Dawkins%20-%20The%20Selfish%20Gene.pdf
17:12 < nsh> muurkha, cool!
17:13 < hprmbridge> Eli> ohhh, "splurge" weed
17:13 < muurkha> sorry, splurge-weed
17:14 < muurkha> Eli: right, shorter telomeres cause you to die sooner to prevent cancer
17:15 < muurkha> most forms of aging can be reasonably explained as ways to prevent cancer
17:16 < hprmbridge> Eli> sorry, I've been speaking a lot of spanish so sometimes my thoughts and words get mixed up. 😵
17:16 < muurkha> no te preocupes, a mí también me pasa bastante
17:19 < hprmbridge> Eli> copado 🫡
17:22 < muurkha> bueno, es una paja, pero qué vas a hacer?
17:27 < hprmbridge> Eli> necessito un mas grande cerebro
17:29 < muurkha> instalate un modelo 70B
17:29 < hprmbridge> Eli> o un neuralink por traducion
17:35 < hprmbridge> Eli> also, my argentina friend gets mad at me for sheismo. So, I'm trying to upgrade my natural language processor
17:47 < fenn> 50M is plenty for a single language pair
17:48 < fenn> wait that can't be right
17:50 -!- mrdata [~mrdata@user/mrdata] has quit [Read error: Connection reset by peer]
17:51 -!- mrdata [~mrdata@135-23-182-55.cpe.pppoe.ca] has joined #hplusroadmap
17:52 -!- mrdata [~mrdata@user/mrdata] has changed host
17:55 < muurkha> language translation is AI-complete
17:56 < muurkha> so bigger models will do it better in the current state of the art
18:01 < fenn> i don't know what that means
18:01 < fenn> i get better translation results with a small model tuned for the task, vs some llama-70B
18:01 < muurkha> .wik AI-complete
18:01 < EmmyNoether> "In the field of artificial intelligence, the most difficult problems are informally known as AI-complete or AI-hard, implying that the difficulty of these computational problems, assuming intelligence is computational, is equivalent to that of solving the central artificial [...]" - https://en.wikipedia.org/wiki/AI-complete
18:03 < fenn> fortunately we have a human brain reading the translated text and doing lots of computation on what it could mean
18:03 < fenn> so the translation doesn't have to be perfect, if there even is such a thing
18:05 < nsh> AI-complete sounds... not rigorous
18:05 < muurkha> it's not, because we don't have a rigorous theory of AI yet
18:08 < nsh> yes well i think the universe was better before Fanya Montalvo took a liberty with this coinage
18:08 < nsh> because it's basically saying "see this thing we have that's lovely? well let's dilute it by applying it to something that necessarily isn't."
18:08 < nsh> and by lovely i mean again rigorous
18:09 < nsh> it's not even aspirational
18:09 < nsh> there's not a conceivable route towards this being equivalent
18:12 < nsh> then again completeness in the context of computational complexity is in a strange halfway house between being potentially meaningful and meaningless
18:12 < nsh> because we have no proven separation between the classes. we just know contingently if they are separate then some problems are equivalent by reducability within the separated classes
18:12 < nsh> so maybe it isn't so far away from AI relative faculty
18:17 < nsh> --
18:17 < nsh> The idea of a Bongard problem is to present two sets of relatively simple diagrams, say A and B. All the diagrams from set A have a common factor or attribute, which is lacking in all the diagrams of set B. The problem is to find, or to formulate, convincingly, the common factor. The problems were popularised by their occurrence in the 1979 book Gödel, Escher, Bach by Douglas Hofstadter, himself a composer of Bongard problems. According to Hofstadter, "the
18:17 < nsh>  skill of solving Bongard problems lies very close to the core of 'pure' intelligence, if there is such a thing".[1] 
18:17 < nsh> -- https://en.wikipedia.org/wiki/Bongard_problem
18:17 < nsh> (lol)
18:22 < muurkha> nsh: her argument is convincing though not rigorous
18:25 < nsh> is it an argument? is it more like
18:25 < nsh> here's a thing
18:25 < nsh> and it does a particular thing that makes it a thing worth calling a thing
18:25 < nsh> and here's something where that doesn't happen
18:25 < nsh> but on a much diminished level, i think some of these things with no mathematical relationship to one another are maybe harder
18:25 < nsh> for reasons
18:26 < nsh> maybe i should see what was argued though :)
18:30 < nsh> i don't think it's even defined in a publication 
18:31 < nsh> like Montalvo isn't in the bibliography for the wp article and papers on scholar.google that mention it only cite her indirectly via the article or someone else who mentioned it
18:31 < nsh> so it's just pub talk imho
18:32 < Ashstar> hm
18:32 < Ashstar> seriously
18:33 < Ashstar> well, yeah, there is a citation bias
18:33 < Ashstar> it seems to spin webs
18:34 < Ashstar> often occluding what is 
18:35 < Ashstar> making more out of what is 
18:36 < muurkha> nsh: yes, there is an argument.  the argument is that people are constantly using knowledge from every aspect of human life to do things like disambiguate visual imagery and disambiguate natural language, and there isn't an easier way to do it, so a computer program will have to be as good as humans are at using knowledge from every aspect of human life before it can do those things as well as humans
18:37 < Ashstar> yup
18:38 < muurkha> https://en.wikipedia.org/wiki/Word-sense_disambiguation#CITEREFBar-Hillel1964 may be the earliest version of this argument
18:38 < muurkha> though lacking the catchy "AI-complete" buzzword of course
18:39 < nsh> hmm
18:40 < Ashstar> yes
18:40 < Ashstar> we're at a point where nformation get cherry picked
18:41 < nsh> https://www.archive.org/download/scientificverse_rg_librivox/scientificverse_11_rankine_128kb.mp3
18:41 < nsh> Three-foot Rule: a Song about Standards of Measure
18:41 < nsh> via: https://librivox.org/a-selection-of-19th-century-scientific-verse/
18:45 < muurkha> because you generally can't preserve ambiguity when you translate between natural languages, you need to perform word-sense disambiguation
18:47 < muurkha> as an example, the "you" in my previous line doesn't refer to nsh, personally, or to a small group of people including nsh, but rather an arbitrarily chosen person faced with this situation.  in many languages you have to use three different words for those different situations.  determining that I meant the third of them involves understanding to what degree the situation I described obtains 
18:48 < muurkha> especially for nsh, or (as in this case) especially *not* for nsh, who does not habitually translate between natural languages
18:48 < muurkha> Ashstar: none of what you said is relevant
18:49 < muurkha> perhaps your natural-language processing is failing to make sense of what is being said
18:49 < Ashstar> we what
18:49 < Ashstar> lol
18:49 < Ashstar> so, ok goo 4 you
18:50 < Ashstar> I just see so much recycling of stuff
18:50 < Ashstar> we have selection biases
18:50 < muurkha> stop talking if you don't have anything to say please
18:51 < Ashstar> go fuck yourself
18:51 < Ashstar> you dissmissive little prick
18:51 < Ashstar> your on idnore
18:51 < Ashstar> fuck you
18:52 < Ashstar> ignoring your stupid ass would be recoomended
18:55 < Ashstar> ok just to be clear, I have worked in the field of both material science (ran a lab at a major U) also, I have a PhD in biology, so muurkha, wtf do you have?
18:55 < Ashstar> hot air?
18:55 < Ashstar> methane owered mouth
18:55 < docl> muurkha is a valued contributor here who consistently makes sound points, we're glad to have him
18:55 < Ashstar> great
18:55 < muurkha> you could go even further than ignoring my stupid ass!  you could actually contribute useful things to the conversation
18:56 < Ashstar> may you all suck up to him
18:56 < docl> he actually talks about material science
18:56 < Ashstar> I ran a fucking alab at Stanford
18:56 < Ashstar> so fuck him'
18:56 < muurkha> yeah, I mean, I could probably talk more sensibly about it if I ran a fucking lab at Stanford
18:56 < docl> from you we got like a few sentences about glass so far
18:56 < docl> signal noise ratio matters
18:57 < Ashstar> I worked on the Hubble
18:57 < Ashstar> so fuck him
18:57 < muurkha> so I wonder why you don't
18:57 < Ashstar> come over
18:57 < docl> I've been going on *nonstop* about mat sci for days
18:57 < Ashstar> let me have my goat do the honors
18:57 < docl> not a peep from the guy who runs a lab
18:58 < muurkha> yeah, unfortunately I haven't had much to offer about the aerographene process, though it's clearly an astounding material
18:58 < Ashstar> you got a great mouth
18:58 < muurkha> and the manufacture process sounds pretty accessible, though not something I can cook up in my kitchen
18:59 < Ashstar> so, sure you can conjour shit out of shit
19:00 < Ashstar> so, either stop being a dick or give those who have worked at this stuff, for a lot longer, probably than you have been alive
19:00 < Ashstar> credit
19:00 < docl> dude, if you want me to read your papers drop a link or something, I'm interested in this stuff. aren't you?
19:00 < muurkha> ditto
19:00 < Ashstar> yes
19:00 < Ashstar> of course
19:01 < Ashstar> just don't need to be insulting others
19:01 -!- test_ [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
19:02 < Ashstar> he has absolutely no idea of what I have done, what I have been involved in
19:02 < docl> of course, you're anon here. why would he know?
19:02 < muurkha> that's because you keep not telling us :)
19:02 < Ashstar> well, to cast aspurions
19:02 < muurkha> I mean Stanford is pretty great, and a PhD is a great achievement, but they aren't replacements for actually participating in a useful way
19:02 < muurkha> it's probably true that I should stop insulting people
19:03 < Ashstar> is just the way thos who are ignorant discredit those who have knowledge
19:03 < Ashstar> aspursions
19:03 < docl> well it's possible to show you have knowledge by explaining stuff
19:03 < muurkha> I think the only time I discredited you was when you confused lux with lumens, and L29Ah got there first
19:03 < Ashstar> n pisses me of, is a bit ignorant
19:04 < muurkha> oh, I'm more ignorant than you can possibly imagine
19:04 < Ashstar> raises my hackles
19:04 < docl> pedantry can be discouraging, but rigor is good, there's tradeoffs in life
19:04 < muurkha> but every day I get a little less so
19:04 -!- flooded [flooded@gateway/vpn/protonvpn/flood/x-43489060] has quit [Ping timeout: 245 seconds]
19:04 < Ashstar> good, sounds like a movie quote
19:04 < Ashstar> oh
19:05 < Ashstar> yes, The Physicician"
19:05 < Ashstar> your quoteing
19:05 < muurkha> really?
19:05 < Ashstar> sounds like it
19:05 < muurkha> I haven't heard of this movie but maybe I should watch it :)
19:05 < Ashstar> good quote
19:05 < muurkha> I thought I was sort of paraphrasing Star Wars
19:05 < Ashstar> excellent movie
19:06 < Ashstar> from the book
19:06 < Ashstar> it is a classic
19:06 < muurkha> https://en.wikipedia.org/wiki/The_Physician_(2013_film)?
19:06 < Ashstar> yep
19:07 < docl> https://www.imdb.com/title/tt2101473/characters/nm2092835
19:08 < docl> same tone, different words "A little better each day"
19:08 < docl> assuming that's the scene
19:09 < Ashstar> yes
19:10 < Ashstar> Kingsley as far as knowing himself
19:10 < muurkha> except that Ibn Sina really *was* a great healer, and I really *am* astoundingly ignorant
19:10 < muurkha> which seems like an important difference
19:10 < Ashstar> he is the Physician
19:10 < Ashstar> well
19:10 < Ashstar> I too admit, I know little
19:11 < muurkha> what recent materials science developments have you found of particular interest?
19:11 < Ashstar> but what was great about working at MIMH, Hubble, Stanford, is that I realize how little I truly know
19:11 < Ashstar> it is humbling
19:11 < docl> don't believe muurkha, he's pretty knowledgable overall (has that consistency seeking trait that's super valuable in a generalist)
19:11 < docl> IMO
19:12 < muurkha> no I'm not
19:12 < Ashstar> that is why it pisses me off
19:12 < Ashstar> to have him so dissmissive
19:12 < docl> see his work here https://dercuano.github.io/
19:12 < muurkha> most of that is wrong
19:12 < muurkha> of Dercuano, I mean
19:13 < docl> see? already iterating :)
19:13 < Ashstar> hm
19:13 < muurkha> well, I was just trying (counterproductively!) to get the discussion back on track
19:14 < Ashstar> just don't be so dissmisive
19:14 < Ashstar> it is insulting, especially when a person is not a "Rube"
19:14 < Ashstar> look it up
19:16 < Ashstar> I have had a long period of some interesting research, a lot of different areas
19:18 < muurkha> what was your favorite?
19:18 < Ashstar> seen stuff - people that are so much further than I can see or achieve
19:19 < Ashstar> I was working with developing cell based biosensors, for neural arrays
19:20 < Ashstar> using my thin film technology
19:21 < muurkha> growing neurons onto an existing substrate to get them to hook up to biocompatible thin-film electrodes?  or more inserting things into cells in situ?
19:21 < Ashstar> then, returned to biotech, I had a start up biotech corp I sold, that produced an antibody binding proetein that sold for $55,000 a gram
19:22 < Ashstar> protein
19:22 < muurkha> it's always nice to get paid
19:22 < Ashstar> which gave me enough to own a winery
19:23 < Ashstar> in Napa
19:23 < docl> you you've done some yeast work I imagine? :)
19:23 < Ashstar> yes
19:24 < Ashstar> even worked with yeast that produce precusors to morphine
19:25 < hprmbridge> fodagut> docl: and they raised the prices of Vyvanse, so the "cheaper" lisdex generic is now the same cash price as Vyvanse was a year or two ago
19:27 < docl> someone needs to make vyvanse yeast
19:30 < hprmbridge> fodagut> lots of people having trouble with the Vyvanse generics, be careful out there. so far I've been lucky
19:31 < docl> so far it seems fine for me. sucks they jacked the price up. I remember the cost for non generic floored me last january, dropped it and moved to adderall for a bit
19:32 < muurkha> docl: there are pseudoephedrine weeds, so it should be feasible
19:32 < hprmbridge> fodagut> well at least for people getting it on insurance, there's usually a different tier (with more contribution by the insurance co) for generic drugs, so for a lot of people the price really did drop
19:32 < hprmbridge> fodagut> but for me I pay the same I did 2-3 years ago, except now I get to play russian roulette with whatever brand they're stocking
19:33 < hprmbridge> fodagut> perhaps literally as I have some sort of hard to diagnose heart issue
19:33 < docl> ugh. no luck with goodrx?
19:35 -!- yashgaroth [~ffffffff@2605:a601:a0cd:9500:78a0:7f36:588:173] has quit [Quit: Leaving]
19:37 < hprmbridge> fodagut> goodrx takes a 30 day supply from $388.77 to $336.50 at my pharmacy. it's better than nothing...
19:40 < docl> https://pubmed.ncbi.nlm.nih.gov/29929980/ // An N-methyltransferase from Ephedra sinica catalyzing the formation of ephedrine and pseudoephedrine enables microbial phenylalkylamine production
19:41 < docl> I'm looking at $100 for the generic. still way too high but less than I remember for January last year
19:42 < muurkha> Ephedra sinica grew wild where I grew up
19:42 < muurkha> they called it "Mormon tea"
19:43 < Ashstar> that's explains Mormons
19:43 < muurkha> the meth craze was just starting to hit, so it was easily available as herbal medicine
19:43 < Ashstar> sex drive
19:43 < docl> I wonder if there are eradication programs it? webmd claims mormon tea is a different plant
19:43 < Ashstar> what a cult
19:44 < docl> Ephedra nevadensis
19:44 < Ashstar> freaking "golden tablets of Moroni"
19:44 < Ashstar> C'mon
19:45 < Ashstar> a cult shit bag of stuff
19:45 < Ashstar> sorry
19:45 < muurkha> heh
19:45 < muurkha> forgiven
19:45 < Ashstar> you mor-mons
19:45 < muurkha> I don't think so, there aren't even eradication programs for cannabis anymore
19:46 < muurkha> maybe in like Yemen or something
19:49 < Ashstar> we restrict some of this stuff, yet these people go down huge amounts of alcohol
19:49 < Ashstar> who regulate it
19:49 < Ashstar> what a farse
19:50 < Ashstar> these compounds are useful
19:50 < docl> apparently they do in my state https://isp.idaho.gov/investigations/programs/
19:50 < Ashstar> especially, some of things, like peyote, psylocibn
19:51 < muurkha> huh, look at that
19:51 < Ashstar> agh
19:51 < Ashstar> what a farse
19:51 < muurkha> and they have a meth lab unit, but no ephedra eradication unit
19:52 < muurkha> pretty sure ephedra grows in Idaho tho
19:52 < docl> ah, good point
19:53 < Ashstar> I hate speed, it robs you of calciulm, causes your nucleus accumbens to be overtaxed
19:54 < Ashstar> shit drug, as far as I am concerned
19:54 < muurkha> well, people like Paul Erdos seemed to do okay on it for decades
19:54 < muurkha> but they weren't taking recreational doses
19:54 < Ashstar> I do think certain stuff like mescaline, psylocipbin is useful, in limted usages
19:55 < Ashstar> marijuana, also
19:55 < Ashstar> I think there is a lot to be gained by their use, MDMA also
19:55 < muurkha> I think marijuana is great at convincing you that it's useful, like alcohol
19:56 < Ashstar> for certain issues
19:56 < muurkha> and I guess it's useful for pain management and nausea management in some cases
19:56 < Ashstar> alcohol is 
19:56 < Ashstar> well, being a vintner
19:56 < Ashstar> awful
19:57 < Ashstar> I started out researching it, at NIMH
19:57 < Ashstar> used to give Rhesuses alcohol
19:57 < muurkha> alcohol is a good low-toxicity organic solvent, a disinfectant, a volatile fuel with some real advantages, and a reasonably good treatment for ethylene glycol poisoning
19:57 < Ashstar> even got drunk with them
19:57 < muurkha> (although there are less toxic treatments for ethylene glycol poisoning available now)
19:57 < Ashstar> I learned drunken rhesus
19:57 < muurkha> beyond that I'm not convinced it has any value
19:58 < muurkha> a friend of mine told me that alcohol was responsible for him getting married
19:58 < Ashstar> alcohol is an organc solvent
19:58 < muurkha> he said he would have been too shy to talk to his then-wife without alcohol
19:58 < muurkha> then she cheated on him and left him
19:58 < muurkha> so I'm not convinced it improved his life either
19:59 < Ashstar> well
19:59 < Ashstar> ok
20:00 < Ashstar> https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6826822/#:~:text=Short%2Dterm%20alcohol%20consumption%20depresses,through%20a%20combination%20of%20both.
20:00 < muurkha> as a fuel it has the advantages that it burns very cleanly, it burns cool enough that you can spill it on your skin without injury (if you put it out fast), and it can be extinguished with water
20:01 < muurkha> and it's far less toxic than most organic liquid fuels
20:01 < muurkha> so it has some real safety advantages as fuel
20:01 < Ashstar> sure, hate to use acetone
20:01 < Ashstar> or diethy ether
20:01 < muurkha> acetone has all of those same advantages actually
20:01 < muurkha> except that it burns a little less cleanly
20:02 < muurkha> and alcohol is an effective food preservative and disinfectant for water or wounds in emergency situations, which cannot be said of most alternative fuels
20:02 < Ashstar> ok
20:02 < muurkha> acetone can actually burn on your skin even longer than alcohol can before it starts burning you
20:03 < Ashstar> still think our livers would be healthier without
20:03 < muurkha> hmm, I'm not sure whether ethanol or acetone has lower hepatic toxicity
20:03 < muurkha> they're both pretty low!
20:03 < docl> it's almost kind of weird we can survive drinking a solvent like this as well as we can
20:04 < muurkha> well, acetone is produced naturally in your body during ketosis
20:04 < docl> I mean, it's pretty different from water
20:04 < muurkha> and ethanol is produced naturally in your gut when you are digesting sugars, sometimes
20:04 < muurkha> and of course is a common contaminant in fruit (even when Ashstar isn't adding it on purpose!)
20:05 < Ashstar> that is a clinical issue
20:06 < docl> https://www.youtube.com/watch?v=0nSI6Afpy1E // Is Alcohol Good for You? | Chris Masterjohn Lite #94
20:06 < muurkha> it's also a major factor in our evolutionary history
20:06 < Ashstar> auto-brewery syndrome (ABS), also known as gut fermentation syndrome, is a very rare disorder. It is characterized by the endogenous production of alcohol. It typically presents with the signs of alcohol intoxication, such as staggering gait, slurred speech, gastrointestinal distress, and state of confusion
20:06 < muurkha> unlike, say, dichloromethane
20:07 < Ashstar> very rare dsorde
20:07 < Ashstar> r
20:07 < muurkha> sure, clinical ABS is very rare, but lower levels of alcohol production are common
20:07 < muurkha> they don't kill us because we're evolved to deal with it
20:07 < Ashstar> not if wee have proper Gi function
20:08 < Ashstar> well, time to go feed my fireplace
20:09 < docl> https://www.youtube.com/watch?v=Qv8pZX7FaT0 // That Mitchell And Webb Look - Slightly Less Than Two Drinks
20:09 < muurkha> consider that a "standard drink" is 14 grams
20:10 -!- supine-gazer [~SupineGaz@2603-6011-89f0-8760-d0bf-8236-a60e-7cfa.res6.spectrum.com] has joined #hplusroadmap
20:10 < muurkha> you could easily produce a gram or two of alcohol with every meal and never know it.  ABS involves producing an order of magnitude more than that
20:12 < docl> yeah. hmm, maybe microdosing would make sense. I bet it's pretty rare for anyone to ingest hormetic quantities on the regular though
20:13 < muurkha> it's common for people to eat fruit that has yeast on it though
20:14 < muurkha> I was trying to find organic solvents that would be deadly if you produced a gram or two of them and I couldn't find any
20:15 < muurkha> (in your gut)
20:16 < muurkha> dichloromethane LD50: 1.25 g/kg.  ethylene glycol: 1 g/kg.  The notorious diethylene glycol: 2-25 g/kg.  Diethyl ether and dimethyl ether: no figures because you can't swallow them liquid.  tetrahydrofuran: 1.7-2.3 g/kg
20:17 < muurkha> hydrazine is 60 mg/kg, so a gram of it somehow being produced in your intestines would stand a chance of killing you, but it's not organic
20:18 < docl> due to no carbon?
20:18 < docl> classification wise, I mean
20:18 < muurkha> yeah
20:18 < docl> ah, ammonia isn't considered organic either
20:19 < muurkha> nope
20:20 < muurkha> it goes on: carbon disulfide: 3.2 g/kg.  carbon tetrachloride: 2-8 g/kg.  perchloroethylene: 2.7-3.4 g/kg.  I guess I'm just not very good at thinking of toxic organic compounds.  except for things like aflatoxin I guess
20:21 < muurkha> hmm, benzene is 0.93 g/kg.  but you still wouldn't die from producing a gram of benzene in your intestines.  maybe you'd get cancer if you did it every day
20:22 < muurkha> melamine, culprit in the fake protein poisoning scandals: 3.9 g/kg
20:22 < docl> apparently one function of glial cells is producing glutamine from ammonia to avoid getting too much ammonia in the brain
20:24 < muurkha> interesting, so if they screw up you can start losing neurons to ammonia?
20:24 < muurkha> wow, even acetonitrile is okay at those levels: 2-2.5 g/kg
20:24 < docl> formaldehyde? I imagine a gram a day of that would be bad
20:24 < muurkha> formaldehyde is a good thought!
20:25 < muurkha> it's 0.1 g/kg.  so the lethal dose for me would be about 10 g.
20:25 < muurkha> if I were producing it in my intestines, though, I'd get progressively worse digestive problems as it cross-linked them
20:25 < docl> wonder how fast it would dilute / get neutralized
20:25 < muurkha> I think it would attack proteins right then and there
20:28 < muurkha> I suspect hydrogen sulfide has a lethal dose close to that of hydrazine
20:28 < docl> well if it hits mostly proteins in your food being digested it might slow the damage rate for gut cells
20:28 < muurkha> yeah. or in bacteria
20:31 -!- TarnyChannie [~TarnyChan@ppp-171-96-101-33.revip8.asianet.co.th] has joined #hplusroadmap
20:32 < docl> https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8879981/ // Unravelling Formaldehyde Metabolism in Bacteria: Road towards Synthetic Methylotrophy
20:32 < TarnyChannie> Hi...
20:32 < docl> howdy
20:33 -!- TarnyChannie [~TarnyChan@ppp-171-96-101-33.revip8.asianet.co.th] has quit [Client Quit]
20:33 -!- TarnyChannie [~TarnyChan@ppp-171-96-101-33.revip8.asianet.co.th] has joined #hplusroadmap
20:34 < TarnyChannie> I have a strategy for what I want to do but... well... thinking about it...
20:35 < TarnyChannie> Would this strategy really be good when in the future you would be 10x faster, 2x more intelligence, have better AI tools, and connect with others better?
20:35 < TarnyChannie> Hi...
20:35 < TarnyChannie> Hmm?
20:36 < docl> sometimes you have to do what you can with what you have. what's your idea?
20:36 < TarnyChannie> I mean... it's a long term strategy... virtual world, whales in the sky, swimming in floating bubbles.
20:36 < TarnyChannie> Except... well... if I decide to program and make a plan now...
20:37 < TarnyChannie> It would likely be biased in a sense that I assume I would be at this intelligence level.
20:37 < TarnyChannie> So I really wonder what to do.
20:37 < muurkha> maybe the best thing to do is to develop your capacities toward being able to achieve your plan, but also other things if you change your mind.
20:38 < muurkha> https://paulgraham.com/hs.html talks about this strategy
20:38 < muurkha> > I think the solution is to work in the other direction. Instead of working back from a goal, work forward from promising situations. This is what most successful people actually do anyway.
20:38 < muurkha> > In the graduation-speech approach, you decide where you want to be in twenty years, and then ask: what should I do now to get there? I propose instead that you don't commit to anything in the future, but just look at the options available now, and choose those that will give you the most promising range of options afterward.
20:38 < muurkha> > It's not so important what you work on, so long as you're not wasting your time. Work on things that interest you and increase your options, and worry later about which you'll take.
20:40 < muurkha> are you in Thailand?
20:41 < TarnyChannie> What?
20:42 < TarnyChannie> Out of all the countries, why do you guess such a specific country?
20:43 < docl> it's showing a .th domain in the join message
20:43 < docl> .m https://twitter.com/eshear/status/1732849060321669188
20:43 < AugustaAva> ​twitter: <eshear> You can beat material into your memory with raw repetition and discipline of course. But it takes so many more reps and so much more energy than the expedient trick of giving a shit about the material instead.
20:44 < docl> fits with what paul graham says "It's not so important what you work on, so long as you're not wasting your time. Work on things that interest you and increase your options, and worry later about which you'll take."
20:44 -!- mxz_ [~mxz@user/mxz] has joined #hplusroadmap
20:45 < TarnyChannie> Oh, I see.
20:45 -!- mxz [~mxz@user/mxz] has quit [Ping timeout: 260 seconds]
20:45 -!- mxz_ is now known as mxz
20:45 < docl> I'm currently following an antimatter rabbit trail that happens to need a lot of aerographite, so now I want to try and make some in my garage
20:47 < docl> but if I accomplish that and develop a scalable method, I could sell aerographite. the stuff is expensive (and useful). could use that to fund the space based solar antimatter collider 
20:47 < TarnyChannie> Transhumanists live around the world.
20:47 < TarnyChannie> I still don't know where I'll live in the future.
20:48 < docl> yeah it's a global community
20:48 < muurkha> docl: even if you don't, you'll surely develop skills in working with difficult materials
20:49 < docl> yep
20:51 < TarnyChannie> I actually had an idea that video games are very efficient sources of happiness and satisfaction, $50 for 1000 hours. So, what about we invest in it like no tomorrow? But still, it's complicated and I ran into some hard walls.
20:52 < TarnyChannie> It was supposed to be a long term plan.
20:52 < TarnyChannie> And then I found out that my ambitious idea would be so ambitious that to execute it, that would take too long, and by that time, the plan would likely change...
20:52 < TarnyChannie> So, I need to find a new plan.
20:54 < docl> educational games or ones that facilitate scientific research always seemed like a good idea to me
20:55 < TarnyChannie> But wait... you're gonna sell aerographite? TBH... I'm a bit focused in the world of bits and digital technologies and perhaps the bottleneck has moved elsewhere. At this point with AI, I'm starting to believe the brain AI which does simple tasks would be more likely done before the muscles. Meanwhile, in the digital realm including coding, art,
20:55 < TarnyChannie> etc... well..
20:55 < docl> pure entertainment games seems like there's already a lot of them, probably only ideal to go into that if it's your driving passion or you can make a lot of money
20:55 < TarnyChannie> AI would likely be better at most jobs within the next 10 years.
20:56 < docl> we shall see about that. also maybe it's our last chance to do anything useful on our own
20:57 < TarnyChannie> Well, "pure entertainment" games aren't necessarily bad. I mean... you could play games or you could try to conquer the world, but the latter option doesn't actually look morally superior given what the world leaders have been doing.
20:57 < TarnyChannie> And well... I'm not saying we'd be out.
20:57 < docl> people are currently selling small samples of aerographene for $200 so it might be pretty lucrative to make cubic meter sized samples
20:57 < TarnyChannie> Ingenuity exists...
20:57 < TarnyChannie> I think we won't go away.
20:58 < TarnyChannie> But we might want to be connected to the brain computer interface.
20:58 < TarnyChannie> This could homogenize intelligence... maybe at least knowledge.
20:58 < TarnyChannie> But it could also amplify intelligence in some forms.
20:59 < TarnyChannie> But well, what I'm sure is that personal preferences remain unique.
20:59 < docl> well the thing that makes games so fun is the learning experience IMO. the designer sets it up so you learn a huge amount in a short while. it's unfortunately mostly not very transferrable to other tasks
21:00 < TarnyChannie> It's actually quite transferable if you play the right games in an abstract sense. And maybe one day games will pose some computational challenges.
21:01 < TarnyChannie> Well, what I've learned from games the most is this...
21:01 < TarnyChannie> Don't assume the world is a balanced game.
21:02 < docl> yeah that's the big difference, the world isn't an optimized learning/playing experience. not at present anyway
21:03 < TarnyChannie> I think Spiffing Brit taught me a lot about the shape of reality.
21:03 < TarnyChannie> If the games where designers tried sooo... hard to be balanced are full of unbalanced mechanics
21:03 < TarnyChannie> There are likely lots of real life mechanics we can exploit.
21:03 < docl> well, there are balances in life you can learn about if you pay attention
21:04 < docl> mechanics that can be exploited
21:04 < fenn> dumb automation is OP
21:06 < fenn> an actually working open source hardware repository https://maslowcommunitygarden.org/
21:06 < fenn> the bootstrappy stuff is under the Maslow tab
21:06 < TarnyChannie> Dumb automation is actually quite good. Intelligence requires extra hardware. Automation should be as dumb as possible while completing the tasks.
21:17 < fenn> compare https://www.youtube.com/watch?v=woH-m3TRakA&t=45s vs https://www.youtube.com/watch?v=mPdviIiNAM8
21:19 < fenn> "professional nail makers could make 100 nails per hour, up to about 4 in a minute, which is insane"
21:20 < fenn> (using mechanized hammers)
21:21 -!- TarnyChannie [~TarnyChan@ppp-171-96-101-33.revip8.asianet.co.th] has quit [Ping timeout: 250 seconds]
21:33 -!- flooded [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
21:37 -!- test_ [flooded@gateway/vpn/protonvpn/flood/x-43489060] has quit [Ping timeout: 264 seconds]
21:45 < muurkha> fenn: it's great to see these Maslow designs!
21:54 < muurkha> fenn: with respect to the nail-making machines, keep in mind that it takes very substantially more labor to make the nail-making machine than it does to make a hammer and anvil
21:54 < muurkha> at least the one in that video
21:54 < muurkha> so it took quite a while for them to become economical
21:56 < fenn> bongard problems are a lot easier having played zendo, otherwise i don't think i'd get the point of it, and would be seen as stupid because the rule i'm supposed to discover was stupid so i wouldn't have guessed something so boring and stupid in an intelligence test
21:56 < muurkha> due to reliability, pairing, etc.
21:56 < fenn> also it's funny naming it bongard problems since bongard credits someone else with having invented it
21:57 < muurkha> wotzisname's law of nomenclature
21:58 < fenn> the nail making machine shown is very highly optimized, yes, but even much less sophisticated machines could outperform a blacksmith
21:58 -!- Gooberpatrol66 [~Gooberpat@user/gooberpatrol66] has joined #hplusroadmap
21:58 < fenn> or any machine at all, such as a trip hammer
21:58 < fenn> that's more augmentation than automation though
21:59 < fenn> what do you mean "pairing"?
22:00 < fenn> also the bongard problem set first leads you through the ontology of the test maker. if presented the problems in a different order it would be much harder
22:01 < fenn> he's teaching the rules of the game at the same time as doing the test
22:02 < fenn> some of these rules like the bounding box of the shape (i guess that's the rule anyway) seem questionably universal
22:03 < fenn> at least in zendo there's a fixed ontology
22:04 -!- supine-gazer [~SupineGaz@2603-6011-89f0-8760-d0bf-8236-a60e-7cfa.res6.spectrum.com] has quit [Read error: Connection reset by peer]
22:14 < muurkha> I have no idea why I said "pairing".  I meant "repair"!
22:17 < fenn> even if the machine is offline 90% of the time it's more productive
22:17 < fenn> actually we can calculate
22:20 < fenn> i count by ear 36 nails produced in 10 seconds by the machine, that's 13000/hr vs 100/hr, so the machine could be offline 99% of the time and still be as productive
22:20 < fenn> they say it does 25800/hr in the description
22:22 < fenn> conversely, we could have 100 dudes working on the machine, or 99 dudes gainfully employed elsewhere in the economy, for the same demand for nails
22:24 < fenn> and 100 nails per hour is quite impressive for a blacksmith in the first place
22:37 < fenn> good luck deciphering this message without the text-based explanation https://www.nsa.gov/portals/75/documents/news-features/declassified-documents/cryptologic-spectrum/communications_with_extraterrestrial.pdf
22:37 < fenn> (page 3)
22:38 < fenn> by comparison the bongard sets are a straightforward and redundant dictionary
22:39 -!- justanotheruser [~justanoth@gateway/tor-sasl/justanotheruser] has quit [Remote host closed the connection]
22:40 -!- justanotheruser [~justanoth@gateway/tor-sasl/justanotheruser] has joined #hplusroadmap
22:50 -!- nmz787_ [~nmz787@user/nmz787] has joined #hplusroadmap
22:51 < nmz787_> fenn: I got the EPC RFID tags and tag reader working using this Python library https://github.com/hex-in/libmagicrf/
22:52 < nmz787_> distance is only 3-4 feet though
22:53 < nmz787_> using this module which, sigh, says 6M (presumably they meant meters) https://www.ebay.com/itm/324103179564
22:53 < nmz787_> .title
22:54 < EmmyNoether> 0-6M Long Range Integrated UHF Reader Writer Module Uart TTL ISO18000 6C  | eBay
22:54 < nmz787_> possibly some messing around with settings will get me further
22:55 < nmz787_> oh, not sure what this fully indicates but "Notice:Using cr80 9662 standard card to test the farthest recognition distance can reach 6M, which is only for reference."
22:56 < nmz787_> these are the tags I bought https://www.ebay.com/itm/154935463481
22:56 < nmz787_> .title
22:56 < EmmyNoether> 50 pcs RFID UHF Passive Smart Label Tag Sticker Antenna Alien 9640 H3  | eBay
22:57 < fenn> have you tried adding a larger ground plane to increase the directionality and range
22:57 < fenn> mount the module on a flat sheet of metal
22:59 < fenn> it does seem kinda like false advertising
22:59 < nmz787_> at the moment it's just connected with a 6" length of ribbon cable jumper wires, from the included (but not listed on ebay) USB-to-serial module
22:59 < nmz787_> so not mounted to anything at all, just sitting on my desk
23:00 < nmz787_> I'll try with some sheet metal
23:00 < fenn> maybe cr80 9662 is an active tag
23:00 < nmz787_> would a curved bowl or satellite dish work better?
23:00 < nmz787_> in addition to the USB 2 serial adapter, it came with some "white card" that shows up in the terminal when I wave it by
23:01 < nmz787_> I just checked, and it seems to get maybe double the range
23:01 < nmz787_> unfortunately, since it wasn't listed on the sale... I am not sure I can figure out what kind it is
23:01 < nmz787_> maybe something I can query
23:02 < fenn> i'd guess it's a cr80 9662
23:02 < fenn> cr80 seems to be the form factor (credit card) and 9662 is the antenna design (squiggly thing with two squares)
23:02 < nmz787_> google seems to concur with photos of a white credit card sized thing
23:02 < nmz787_> ah
23:04 < nmz787_> .gpt 9640 vs 9662 tag
23:04 < gptpaste> ​The tags "9640" and "9662" do not have any specific meaning or context provided. Therefore, it is difficult to determine the difference between them without further information.# Wed 27 Dec 08:04:16 CET 2023 - 9640 vs 9662 tag - https://bin.gy/luggeresse
23:04 < fenn> .gpt4 9640 vs 9662 tag
23:04 < nmz787_> .gpt what is the difference between the RFID tags 9640 and 9662?
23:04 < nmz787_> oops
23:05 < nmz787_> I think I confused the bot
23:05 < gptpaste> ​It seems like you are referring to product or model numbers, which could be associated with various items depending on the context. Without additional information, it's difficult to provide a specific answer.If these tags are related to technology products, they might denote different versions or models of a device where "9640" and "9662" could indicate slight variations  - https://bin.gy/krydectaly
23:05 < gptpaste> ​The main difference between RFID tags 9640 and 9662 lies in their respective features and capabilities. Here are some key distinctions:1. Frequency: The RFID tag 9640 operates at a frequency of UHF (Ultra High Frequency), typically around 860-960 MHz, while the RFID tag 9662 operates at HF (High Frequency), usually around 13.56 MHz.2. Read Range: The read range of the RF - https://bin.gy/agunduante
23:05 < fenn> i really thought gpt4 would have figured that one out on its own
23:05 < nmz787_> also I don't think this reader can do 12MHz
23:06 < nmz787_> or that order of magnitude anyway
23:06 < fenn> it sure looks like 9640 is the right tag for the listed frequency range
23:07 < nmz787_> 9640 - https://www.alientechnology.com/products/tags/squiggle/
23:07 < nmz787_> 9662 - https://www.alientechnology.com/products/tags/short/
23:07 < fenn> These products are covered by one or more of the following U.S. patents: 7967204, 7931063, 7868766, 7737825, 7716208, 7716160, 7688206, 7659822, 7619531, 7615479, 7598867, 7580378, 7576656, 7562083, 7561221, 7559486, 7559131, 7554451, 7551141, 7542301, 7542008, 7531218, 7522055, 7500610, 7489248, 7453705, 7425467, 7417306, 7411503, 7385284, 7377445, 7364084, 7353598, 7342490, 7324061, 7321159, 7301458, 
23:07 < fenn> 7295114, 7288432, 7265675, 7262686, 7260882, 7253735, 7244326, 7218527, 7214569, 7199527, 7193504, 7173528, 7172910, 7172789, 7141176, 7113250, 7101502, 7080444, 7070851, 7068224, 7046328, 6998644, 6988667, 6985361, 6980184, 6970219, 6952157, 6942155, 6933848, 6927085, 6816380, 6780696, 6731353, 6693384, 6683663, 6665044, 6657289, 6623579, 6606247, 6606079, 6590346, 6586338, 6566744, 6555408, 6527964, 
23:07 < fenn> 6479395, 6468638, 6420266, 6316278, 6291896, 6281038. Other patents pending.
23:08 < fenn> lmao
23:10 < nmz787_> also, I'm not sure what the difference is or why anyone would care, other than regulators or something... but I received a European freq range reader, though I ordered the USA freq range reader
23:10 < nmz787_> the tags seem to not care
23:11 < fenn> clearly black helicopters will be descending upon your rural homestead shortly, to recover the lost alien technology
23:12  * nmz787_ slowly lowers the window blinds
23:13 < fenn> .gpt4 what are NVAM bits in the context of RFID tags
23:14 < gptpaste> ​NVAM stands for Non-Volatile Auxiliary Memory, which refers to a type of memory that can retain stored information even when not powered. In the context of RFID (Radio-Frequency Identification) tags, NVAM bits would be part of the non-volatile memory where additional data can be stored.RFID tags typically contain a microchip with memory and an antenna for communication. T - https://bin.gy/mulsotmans
23:15 < nmz787_> hmm, I guess that's so you don't need an external database to correlate the tag ID with "ryobi drill"
23:15 < fenn> there are lots of ways to use it
23:16 < muurkha> fenn: yes, being offline isn't what makes the cost of repair important.  what makes the cost of repair important is if it takes you more to repair the machine than it would have taken you to manufacture the nails it autofactures before it breaks again
23:16 < fenn> you can log temperatures, gps waypoints, timestamps, etc
23:16 < muurkha> the number for the blacksmith was 240/hour
23:16 < fenn> a heavily mechanically augmented blacksmith
23:17 < muurkha> maybe, it's not clear whether that's with a trip hammer or without
23:17 < fenn> with
23:17 < muurkha> it's not clear
23:17 < fenn> either both numbers are with a trip hammer or just 240/hr
23:18 < fenn> i guess i could go back and listen to the guy ramble, but i have this book of bongard problems to stare at
23:18 < muurkha> the blacksmith who is reporting the 240 nails per hour number has never seen it done.  he's just reporting hearsay
23:21 < muurkha> enoy
23:21 < muurkha> enjoy
23:22 < fenn> nmz787_: the GUI shows a button to change the RF power
23:22 < nmz787_> yeah, I've tried some stuff in the python
23:22 < nmz787_> READER = m100(SERIAL_PORT)
23:22 < nmz787_> READER.region(CONSTANTS.REGION_EUROPE)
23:22 < nmz787_> READER.demodulator(mixer=CONSTANTS.MIXER_GAIN_16DB, ifgain=CONSTANTS.IF_GAIN_40DB)
23:22 < nmz787_> READER.mode(CONSTANTS.MODE_HIGH_SENSITIVITY)
23:23 < nmz787_> from another github repo, it seems like I might also need to set the specific channel I want to use, after setting the region
23:23 < nmz787_> (seems like one would imply the other, but, who am I to say)
23:24 < hprmbridge> godmode7777> Oh man Thats the exact group i wanted. I always thought I was the only one who thinks like crazy about these stuff. glad to join.
23:27 < nmz787_> oh, welp, looks like this python module doesn't have an implementation for setting the channel :/ https://github.com/hex-in/libmagicrf/blob/master/src/lib/_protocol.c#L926
23:29 < fenn> come on at least raise a NotImplementedError
23:31 < fenn> nmz787_: can you see the returned power of the tags?
23:32 < nmz787_> how do the kids these days find questionable chinese software titled "uhf rfid reader app v2.1" ?
23:32 < nmz787_> I see -## dBm
23:32 < nmz787_> 2023-12-26 23:06:50 -> E280699500007003A2F3FCD1 RSSI: -48 dBm
23:32 < nmz787_> 2023-12-26 23:06:50 -> E280699500007003A2F3FCD1 RSSI: -45 dBm
23:32 < fenn> email the seller
23:32 < fenn> or "ask a question" in ebay parlance
23:33 < fenn> hmm so maybe you'll have to average a bunch of reads to get some reasonable precision
23:34 < nmz787_> I *feel* like the module is already doing that
23:34 < nmz787_> based on something I half-ass read earlier
23:35 < fenn> dBm(-45)/dBm(-48) = 2
23:36 < nmz787_> heh, well the seller already replied to my statement about me getting the wrong region "Hello friend, we will resend a module to you as soon as possible. ---sorry for the miastake."
23:38 < fenn> well if you have to end up scanning some dish around to get the necessary range, at least that will give you directionality
23:38 < fenn> kinda disappointing though
23:39 < nmz787_> well I was already assuming I'd be lugging around some sort of magic wand while looking for some tool that I can't find
23:39 < fenn> it could still be used for "when did this tool leave through a door" tracking, by putting the reader module on both exits
23:39 < nmz787_> ah, yeah
23:42 < fenn> who's running on the "a phased array on every rooftop" platform
23:42 < nmz787_> I am thinking have some sort of small touchscreen, where I can enter characters and they are used to substring match from the known database, and then a particular tag can be selected (or multiples). Then when that/those tag(s) come into range, the thing buzzes or beeps... maybe changing pitch as it gets closer
23:43 < fenn> voice control? can whisper run on a pi yet?
23:43 < nmz787_> it seems the smart phone would be the best I/O... but then I need to deal with compiling an android app, including bluetooth, and then getting something like an esp or esp32 for the reader side of bluetooth\
23:44 < fenn> can android talk directly to the modem over the USB-serial dongle?
23:44 < nmz787_> hmm, yeah I suppose that could work
23:45 < nmz787_> unless they've crappified the OS so much that you need root, etc
23:45 < fenn> i have no idea
23:45 < fenn> last i looked into this was getting keyboards and sound cards on the nook, which was android 2.1
23:46 < fenn> usb-otg was always a crapshoot
23:46 < fenn> if the esp32 can run a webserver that would simplify everything
23:47 < nmz787_> yeah they can
23:47 < fenn> stupid android, making us use even more webservers
23:47 < fenn> i hope 2024 is peak web server
23:48 < fenn> did you ever try ad-hoc wifi between a phone and an esp32?
23:49 < fenn> to talk to your trees for example
23:49 < nmz787_> no, I haven't played with the esp32 much... I got an esp32-cam module setup with some demo code that logs onto my house wifi and works to stream video.... and that's about it
23:50 < fenn> ok but then you need a wifi access point
23:50 < nmz787_> I want more webcams sooner than later... for example one staring at the compost pile might prove interesting as a months long time-lapse
23:51 < nmz787_> the demo code I used was through Arduino, so I'm pretty sure they must have ad-hoc mode demos
23:54 < fenn> there is an app called serial bluetooth terminal, and it just sends and receives text
23:56 < nmz787_> even if the USB to serial adapter "just works" in the sense that it shows up in /dev/ttyUSB##  there's still the actual program to encode+send and receive+decode the packets from the reader, as well as the UI stuff
23:56 < nmz787_> this Python library with it's C-extension... I just have no idea how much work that would be to get running
23:56 < nmz787_> (while on my laptop it was just "python setup.py install"
23:57 < nmz787_> (ok I had to google for some #define that needed added since my version of Python is so new it added some extra crap for C extensions)
23:57 < fenn> if you put all the UI stuff in the ESP32 code, you don't have to worry about android development
23:57 < nmz787_> hmm
23:57 < fenn> think 1980s library card catalog interface
23:58 < fenn> or something even less fancy
23:58 < nmz787_> yeah, that's true... I believe I have seen MicroPython for esp32
23:59 < nmz787_> though at some point I guess I'm just trying to be lazy and not re-implement the low-level reader protocol
23:59 < nmz787_> which as I found out, is incomplete
23:59 < nmz787_> it /is/ C code though, so maybe not too hard to shove that into the esp32 arduino demo code
--- Log closed Wed Dec 27 00:00:11 2023