--- Log opened Thu Jul 11 00:00:49 2024
00:08 -!- gl00ten [~gl00ten@bl11-102-149.dsl.telepac.pt] has joined #hplusroadmap
00:16 < TMA> jrayhawk: I think the consensus is that it is OK to train LLMs on copyrighted material, maybe the epistemic tree could be bundled unde this as well
00:18 < jrayhawk> that's a funny thought
00:20 < TMA> maybe not. maybe it falls under the category of "we can, but jrayhawk cannot" to paraphrase my cousin argumenting that our grandfather forbidding to do something does not apply to them
00:23 < jrayhawk> copyright only protects the specific expression of an idea, not the idea itself. "transformers" are almost always sufficiently transformative that the original expression of the idea is effectively gone
00:24 < jrayhawk> epistemology oracles are great, though. i get to ask islamobot how confident it is that Adam was 30 meters tall.
00:26 < jrayhawk> epistemology oracles are great, though. i get to ask islamobot how confident it is that Adam was 30 meters tall.o
00:26 < jrayhawk> whoops
00:27 < jrayhawk> i.e. i can find out from a variety of commonly accepted epitemic weightings how truthy something is
00:33 < jrayhawk> as opposed to now, where the editor most stubbornly camping the article over the past year selects for stale citations of passages in papers and books from years past that they haven't read and assume nobody else will be able to, either
01:28 -!- gl00ten2 [~gl00ten@2001:8a0:7ef6:8700:19b7:5269:6d30:a6bf] has joined #hplusroadmap
01:31 -!- gl00ten [~gl00ten@bl11-102-149.dsl.telepac.pt] has quit [Ping timeout: 256 seconds]
01:35 < jrayhawk> "Meta-Bot, what specific oracle bots would best represent the epistemic weighting system of me and/or my political opponent?" "Specific-Oracle-Bot, how confident are you in prediction x? What priors most heavily weight the strength of your prediction? What are the lowest-accessibility priors in that list?"
02:41 -!- gl00ten2 [~gl00ten@2001:8a0:7ef6:8700:19b7:5269:6d30:a6bf] has quit [Remote host closed the connection]
02:42 -!- gl00ten [~gl00ten@2001:8a0:7ef6:8700:19b7:5269:6d30:a6bf] has joined #hplusroadmap
03:19 -!- darsie [~darsie@84-112-12-36.cable.dynamic.surfer.at] has quit [Remote host closed the connection]
03:19 -!- darsie [~darsie@84-112-12-36.cable.dynamic.surfer.at] has joined #hplusroadmap
03:30 -!- darsie [~darsie@84-112-12-36.cable.dynamic.surfer.at] has quit [Remote host closed the connection]
03:31 -!- darsie [~darsie@84-112-12-36.cable.dynamic.surfer.at] has joined #hplusroadmap
04:16 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has quit [Read error: Connection timed out]
04:23 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has joined #hplusroadmap
05:28 -!- darsie [~darsie@84-112-12-36.cable.dynamic.surfer.at] has quit [Remote host closed the connection]
05:29 -!- darsie [~darsie@84-112-12-36.cable.dynamic.surfer.at] has joined #hplusroadmap
05:50 < hprmbridge> Eli> @jrayhawk I’m getting every blood test from that link fenn posted for cheap blood tests. If my ApoB is high and I have low inflammation, what would you do? Is tnf alpha/il6 measurement necessary, or is just hsCRP acceptable, in your opinion?
05:50 < hprmbridge> Eli>
05:50 < hprmbridge> Eli> The studies (AMORIS for example) suggest that extremely low ApoB is highly correlated with low ASCVD. So even if ApoB is a subpar biomarker, if I understand your argument correctly, it is still a legit biomarker according to studies, and seems to be one that is easily reduced. If I can get a super reduced price on the PCSK9 inhibitor I’ll do that. Otherwise I’ll probably go with like a
05:50 < hprmbridge> Eli> rosuvastatin ezetimibe combo since it’s so dang cheap.
06:14 < nsh> my advice would be to walk five miles between every consideration of a biomarker
06:14 < nsh> or daily, whichever works out as more walking
06:15 < nsh> if the routes are particularly monotonous you might even find you can bikeshed your health while walking them
06:15 < nsh> but the walking will do better for it almost certainly
06:16 < nsh> unless you're about to die from something you find out about through such fascinations
06:19 < hprmbridge> Eli> I already exercise. I think that’s not enough though. Not just because have high BP.
06:19 < hprmbridge> Eli>
06:19 < hprmbridge> Eli> ASCVD is the top killer in America. The hot topic rn is how much we can reduce it. If everyone had good lpa levels and ApoB of 30mg dL some doctors now think it could potentially even reduce ASCVD to an orphan disease status. I have no idea of the truth of that. But people I respect have been cautiously optimistic about how we could knock ASCVD from the top killer spot.
06:25 < hprmbridge> Eli> Perhaps it’s just another fad that will be considered harmful in the future (like prolonged fasting and metformin). I don’t have any idea. I think we can say statins are some of the most well studied exogenous molecules so far. Perhaps up there with alcohol studies done every weekend in colleges
06:29 -!- gl00ten [~gl00ten@2001:8a0:7ef6:8700:19b7:5269:6d30:a6bf] has quit [Ping timeout: 268 seconds]
06:32 < TMA> Eli: you measure X that is a marker for Y. If you go on optimizing X, you'll just destroy the correlation to Y (for average X and Y), because there might be some Q, R, S that influence the value of X and are influenced by Y and Z
06:33 < TMA> you might be tweaking Z, which drowns out any influence Y has on X via Q,R,S...
06:34 < TMA> you need much more knowledge of the X-Y relation to be able to influence Y by attempting to tweak X
06:54 < hprmbridge> Eli> It’s true. The human body is a complex dynamic system. The only thing certain is that determining anything with absolute certainty is impossible. Having said that, we have over 30 years of data with a few million people. Since it’s the biggest killer it’s been studied as well as any exogenous molecules I can think of. The current data is pointing to reducing atherosclerosis via a reduction in ApoB
06:54 < hprmbridge> Eli> early on. Ideally they will test that next but the trial would require decades of study. No one needs to follow me. I’ve been wrong before. I’ll probably be wrong in the future.
07:04 < nsh> i thought everyone here was on a 10 year break from optimising things
07:05 < nsh> :P
07:05 < TMA> nsh: 10 year break from optimising things == 10 years dead
07:06 < nsh> almost every living thing i can think of has never considered a single variable
07:07 < nsh> one notably exceptional species however has done so and hardly covered itself in glory so doing
07:07 < TMA> almost every living thing is not conscious in a way permitting such consideration
07:07 < TMA> it does not mean they don't optimize
07:08 < TMA> plants optimize their growth direction towards sunlight, water and nutrients
07:09 < TMA> by perusing gradient descent optimization technique
07:09 < nsh> .ety peruse
07:09 < saxo> "peruse (v.)late 15c., 'to go through searchingly or in detail, run over with careful scrutiny,' from Middle English per- 'completely' (see per) + use (v.). Meaning 'read carefully and critically' is by 1530s, but this could be a separate formation. Meaning 'read casually' is from 19c. Related: Perused; perusing. '..." - https://www.etymonline.com/search?q=peruse
07:10 < nsh> the plants do not, the ecosphere by a distribute evolutionary process can be thought of so to do
07:10 < nsh> *distributed
07:10 < nsh> although scrutiny here is being stretched perhaps to breaking point
07:10 < nsh> we can of course project what we have taken it upon ourselves to do as part of scientific endeavour back onto plants
07:11 < nsh> but it's only a tiny bit less facile than saying a floating leaf is optimising its depth relative to its weight and buoyancy
07:11 < TMA> nsh: grab a pea. let it germinate. put it in a place that permits light just from one small opening. observe how it will grow towards that opening
07:11 < hprmbridge> kanzure> https://nymag.com/intelligencer/article/adderall-shortage-adhd-medication-ascent-pharmaceuticals.html
07:12 < nsh> TMA, thanks for reminding me of what cress and mustard were kind enough to show me when i was six or seven :)
07:12 < nsh> even then i could have told you that they weren't considered direction nor quantity of radiative flux
07:13 < nsh> perhaps not those latter words
07:13 < nsh> *considering
07:13 < nsh> anyway this isn't the point
07:14 < nsh> the point is Certain People make a farce out of considering variables and measuring them and optimising some
07:14 < nsh> and they'd do well to take a rest from it :)
07:14 < nsh> and take up a racket sport maybe
07:14 < nsh> or holy orders
07:21 < TMA> that's hair splitting. they have not _thought_ about that. but they considered it as in _reacted_ to that. in the meaning of consider == don't ignore
07:21 < TMA> but I agree with your restated point
07:41 -!- darsie [~darsie@84-112-12-36.cable.dynamic.surfer.at] has quit [Remote host closed the connection]
07:42 -!- darsie [~darsie@84-112-12-36.cable.dynamic.surfer.at] has joined #hplusroadmap
07:54 -!- darsie [~darsie@84-112-12-36.cable.dynamic.surfer.at] has quit [Remote host closed the connection]
07:54 -!- darsie [~darsie@84-112-12-36.cable.dynamic.surfer.at] has joined #hplusroadmap
08:24 -!- juri_ [~juri@implicitcad.org] has quit [Ping timeout: 252 seconds]
08:30 -!- darsie [~darsie@84-112-12-36.cable.dynamic.surfer.at] has quit [Remote host closed the connection]
08:31 -!- darsie [~darsie@84-112-12-36.cable.dynamic.surfer.at] has joined #hplusroadmap
08:34 -!- etc-vi [~etc-vi@user/meow/girlchunks] has quit [Quit: nya]
08:35 -!- etc-vi [~etc-vi@user/meow/girlchunks] has joined #hplusroadmap
08:36 -!- juri_ [~juri@implicitcad.org] has joined #hplusroadmap
08:39 -!- etc-vi [~etc-vi@user/meow/girlchunks] has quit [Client Quit]
08:40 -!- etc-vi [~etc-vi@user/meow/girlchunks] has joined #hplusroadmap
08:43 -!- gl00ten [~gl00ten@2001:8a0:7ef6:8700:19b7:5269:6d30:a6bf] has joined #hplusroadmap
08:50 -!- gl00ten [~gl00ten@2001:8a0:7ef6:8700:19b7:5269:6d30:a6bf] has quit [Ping timeout: 240 seconds]
08:57 < geneh2> adderall factory in a box that takes in air and electricity and outputs adderall.
09:00 < geneh2> is it inefficient? absolutely, but it means it's diffifult for someone running it to make more than the dose they need
09:18 -!- Gooberpatrol66 [~Gooberpat@user/gooberpatrol66] has quit [Ping timeout: 268 seconds]
09:22 < hprmbridge> kanzure> is that a thing that exists
09:28 -!- faceface [~faceface@user/faceface] has quit [Quit: Lost terminal]
09:33 < geneh2> it doesn't exist yet, but all the tech is there
09:34 < geneh2> getting pharmaceutical purity may be difficult though. Most of the box would probably be full of stuff to test purity
09:45 -!- darsie [~darsie@84-112-12-36.cable.dynamic.surfer.at] has quit [Remote host closed the connection]
09:46 -!- darsie [~darsie@84-112-12-36.cable.dynamic.surfer.at] has joined #hplusroadmap
10:04 -!- Gooberpatrol66 [~Gooberpat@user/gooberpatrol66] has joined #hplusroadmap
10:22 -!- TMM [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
10:22 -!- TMM [hp@amanda.tmm.cx] has joined #hplusroadmap
10:33 < jrayhawk> Eli: an oxLDL test would be a useful indication of abnormally poor LDL receptor activity in the liver or peripheral tissues, or abnormally high oxidative stress, which gets you slightly closer to meaningful etiology.
10:37 < jrayhawk> CRP and friends are at-least-interesting if they're high, but adhesion and permeability are potentially transient events (for instance, with prolamine, glycoalkaloid, or lectin consumption)
10:43 < jrayhawk> imaging is the thing that is properly diagnostic of atherosclerosis
11:00 < jrayhawk> if LDL-P is high, you need to work out *why* it's high, *if* it's a problem, and *if* the dozen interventions you can make that are likely to be better cost/benefit/harm tradeoffs than statins are sufficient.
11:04 -!- darsie [~darsie@84-112-12-36.cable.dynamic.surfer.at] has quit [Remote host closed the connection]
11:05 -!- darsie [~darsie@84-112-12-36.cable.dynamic.surfer.at] has joined #hplusroadmap
11:06 < jrayhawk> if you have a fourth the LDL receptor expression/activity, having four times the LDL-P concentration is entirely innocuous IF it is not oxidizing.
11:11 < jrayhawk> (although there is a distinction to be drawn between lumenal oxidation (easy to measure) and sub-lumenal oxidation (difficult to measure) that means we also care about adhesion and permeability in some circumstances
11:11 < jrayhawk> +)
11:27 < docl> startup idea: nicotine smart patch that delivers when you commit code
11:28 < docl> (basically taking the hard bits out of my earlier idea of amphetamine delivering implants and combining with operant conditioning)
11:29 < docl> maybe have it beep as it starts the delivery so the brain starts anticipating it immediately
11:30 < nsh> pasky, your claude.vim has been appreciated in ##LLM, you might like to consider joining. (pardon me if it's been suggested already)
11:31 < nsh> just watching https://www.twitch.tv/VictorTaelin who doing vim+sonnet+kind2 live programming atm
11:31 < nsh> which reminded me indirectly
11:32 < nsh> .m https://x.com/VictorTaelin/status/1809290888356729002
11:32 < AugustaAva> ​twitter: <VictorTaelin> Code editing has been deprecated ␤  ␤ I still need to be there to feed insights, but the actual editing is fully automated. I now program by just talking to Sonnet on terminal. This complex refactor should take days, and it was done by lunchtime. How long til it is fully autonomous? https://video.twimg.com/amplify_video/1809289984870084609/vid/avc1/720x720/XWdGyQCvOEeKB45N.mp4?tag=14
11:33 < nsh> (seems they wrote their own plugin using gpt4)
11:33 < nsh> don't like the idea of spending $53 for a day's code refactoring just on LLM API calls however
11:33 < nsh> but i've never really had to refactor a codebase so hard to assess the value of doing it more efficiently
11:42 < jrayhawk> streamlink + streamlink-ttvlol is good for twitch viewing. and there's also a legacy undocumeted anonymous IRC read interface for chat: irc://justinfan56789@irc.twitch.tv:6667/#victortaelin
11:42 < jrayhawk> er, ircs://
11:49 < L29Ah> what is this language?
11:49 < L29Ah> ok, kind2
11:57 -!- darsie [~darsie@84-112-12-36.cable.dynamic.surfer.at] has quit [Remote host closed the connection]
11:58 -!- darsie [~darsie@84-112-12-36.cable.dynamic.surfer.at] has joined #hplusroadmap
12:02 -!- gptpaste [~x@193.29.56.54] has quit [Read error: Connection reset by peer]
12:02 -!- AugustaAva [~x@193.29.56.54] has quit [Remote host closed the connection]
12:05 < hprmbridge> kanzure> docl, we used to just feed programmers m&m candies. seems simpler to me?
12:09 < jrayhawk> chocolate does contain TAAR1 agonist
12:12 < docl> yes but calories / seems harder to automate (I did try this BTW)
12:13 < docl> er, tried to do it manually not automate
12:20 < pasky> nsh: thx! yeah i follow victor, he has a bit different take on it (honestly i got really frustrated with claude today though, i think i need to prompt it *waaay* better, should take a look at how he does it)
12:21 < pasky> nsh: an interesting dynamic is that it still takes some tries, and i'm not sure if i'm actually faster or slower with claude right now; but i'm sure as heck reltively frustrated when claude isn't getting it right on 3rd attempt, it taps into my extroversion energy (which means it's mentally more exhausting than coding on m own)
12:21 < pasky> i hope better prompts should help a lot here though
12:26 -!- gptpaste [~x@193.29.58.204] has joined #hplusroadmap
12:26 < nsh> hmm, hadn't thought of that
12:27 < nsh> (the depletion of social interaction energy when intensively/discursively relying on LLM interactions)
12:28 < nsh> difficulties in such early days to have a good sense of what is passing growing pains and what we should get used to as more intrinsic/inherent qualities of things to  
12:28 < nsh> come
12:28 < pasky> yep
12:28 < pasky> BTW probably state-of-art in this area is aider
12:28 < pasky> just needs much better UX
12:29 < pasky> I wondered already about using it as a backend
12:29 < pasky> another thing is it's actually rather complex instrumentation, not very easy to see what exactly is going on
12:31 -!- gptpaste [~x@193.29.58.204] has quit [Remote host closed the connection]
12:33 -!- darsie [~darsie@84-112-12-36.cable.dynamic.surfer.at] has quit [Remote host closed the connection]
12:34 -!- darsie [~darsie@84-112-12-36.cable.dynamic.surfer.at] has joined #hplusroadmap
12:40 -!- gptpaste [~x@193.29.58.204] has joined #hplusroadmap
12:57 -!- AugustaAva [~x@193.29.58.204] has joined #hplusroadmap
14:08 < docl> https://sarahconstantin.substack.com/p/superbabies-putting-the-pieces-together
14:39 -!- Gooberpatrol66 [~Gooberpat@user/gooberpatrol66] has quit [Ping timeout: 256 seconds]
14:42 -!- darsie [~darsie@84-112-12-36.cable.dynamic.surfer.at] has quit [Ping timeout: 264 seconds]
15:51 -!- Gooberpatrol66 [~Gooberpat@user/gooberpatrol66] has joined #hplusroadmap
16:43 -!- Gooberpatrol66 [~Gooberpat@user/gooberpatrol66] has quit [Ping timeout: 268 seconds]
17:04 -!- Netsplit *.net <-> *.split quits: mxz, flyback, ike8, TMA, SDr, juri_, archels_, superkuh, stipa, kanzure,  (+9 more, use /NETSPLIT to show all of them)
17:06 -!- Netsplit over, joins: saxo
17:07 -!- Netsplit over, joins: kanzure, TMM, etc-vi, juri_, L29Ah, mxz, superkuh, flyback, ike8, TMA (+4 more)
17:09 -!- 029ABLEB0 [~mxz@user/mxz] has joined #hplusroadmap
17:09 -!- SDr [~SDr@user/sdr] has joined #hplusroadmap
17:09 -!- berndj [~berndj@197.189.254.139] has joined #hplusroadmap
17:09 -!- dartmouthed [~blackunsp@li761-35.members.linode.com] has joined #hplusroadmap
17:44 -!- Gooberpatrol66 [~Gooberpat@user/gooberpatrol66] has joined #hplusroadmap
18:07 < hprmbridge> kanzure> "Induction of meiosis from human pluripotent stem cells" https://www.biorxiv.org/content/10.1101/2024.05.31.596483v1
20:17 -!- mxz_ [~mxz@user/mxz] has joined #hplusroadmap
20:17 -!- 029ABLEB0 [~mxz@user/mxz] has quit [Ping timeout: 264 seconds]
20:18 -!- mxz [~mxz@user/mxz] has quit [Ping timeout: 260 seconds]
20:18 -!- mxz_ is now known as mxz
20:24 -!- TMM [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
20:25 -!- TMM [hp@amanda.tmm.cx] has joined #hplusroadmap
21:30 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has quit [Read error: Connection timed out]
22:25 -!- mxz_ [~mxz@user/mxz] has joined #hplusroadmap
22:55 -!- darsie [~darsie@84-112-12-36.cable.dynamic.surfer.at] has joined #hplusroadmap
23:21 < hprmbridge> Eli> Isn’t that what programmers already do with vyvanse and Red Bull?
23:26 < docl> doing it manually is not the same as having it automated for you
23:27 < docl> also vyvanse is designed to be delivered slowly over time to be less habit forming so it's not optimal here
23:29 -!- darsie [~darsie@84-112-12-36.cable.dynamic.surfer.at] has quit [Remote host closed the connection]
23:33 < hprmbridge> Eli> Just have a neuralink plugged in your dopamine system bro
23:33 < hprmbridge> Eli> And download the dopamine module from NPM
23:33 -!- darsie [~darsie@84-112-12-36.cable.dynamic.surfer.at] has joined #hplusroadmap
23:37 < hprmbridge> Eli> I’m assuming the ct angiogram is the state of the art currently?
23:43 < jrayhawk> I think FFRCT is supposed to be? I don't keep up with imaging very well. At least CAC is the cheap high-sensitivty, low-specificity option good for a first pass, and some CT variation is the expensive high-sensitivity-high-specificty test.
23:45 < jrayhawk> https://peterattiamd.com/ethanweiss2/ there's probably some relevant information in here somewhere
--- Log closed Fri Jul 12 00:00:50 2024