--- Log opened Fri Feb 24 00:00:53 2023
00:49 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has quit [Ping timeout: 252 seconds]
02:00 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has joined #hplusroadmap
02:12 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has left #hplusroadmap []
02:35 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has joined #hplusroadmap
02:56 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has quit [Ping timeout: 264 seconds]
03:00 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has joined #hplusroadmap
03:14 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has quit [Ping timeout: 255 seconds]
04:34 < nsh> copyright infringement is when you're smol and not bigly
04:35 < nsh> pretending that it matches some set of objective principles is a game that lawyers play with judges
04:35 < nsh> and charge you a lot of money to watch
04:35 < nsh> which further underpins the original observation
04:48 < hprmbridge> kanzure> jiankui he has been deleting his embryo editing tweets today. like the one about the first donation to the lulu and nana foundation.
04:54 < hprmbridge> kanzure> guess i'll just recite it from memory for posterity https://twitter.com/kanzure/status/1629102263149928450
05:37 < kanzure> "FORCE (first-order reduced and controlled error) learning"
05:52 < kanzure> re: the "compliant mechanisms" and "origami antennas" published earlier this month https://www.nature.com/articles/s41467-023-36143-6 i was wondering whatever became of the fluidic antenna stuff, did that go anywhere
05:53 < kanzure> like this fluidic antenna https://onlinelibrary.wiley.com/doi/10.1002/adfm.200900604 "a fluid metal alloy injected into microfluidic channels comprising a silicone elastomer"
07:14 < nsh> i suspect fluid is a characteristic that improves the characteristic antenna
07:14 < nsh> is *not
07:15 < nsh> perhaps for forming but not during use
07:50 < kanzure> have there been any attempts at large-parameter machine learning models for controllers of human learning? https://www.sciencedirect.com/science/article/pii/S0896627309005479
07:51 < kanzure> in particular figure 1c (not the best diagram but it shows the concept)- there are weights used outside the biological network and get adjusted to produce the output value z, and feedback is used to minimize the error
07:53 < kanzure> (in particular one application would be using it to teach a person some arithmetic but without using working memory numerical manipulation, or a simpler problem)
08:51 -!- cthlolo [~lorogue@77.33.23.154] has joined #hplusroadmap
09:24 < muurkha> fenn: usually there are a number of different rights that copyright reserves to the copyright holder; sale is one of them, and probably what you're thinking about, but others usually include public performance, making copies, translation, synchronizing with video, etc., though it depends on the country
10:22 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has joined #hplusroadmap
11:03 -!- TMM_ [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
11:03 -!- TMM_ [hp@amanda.tmm.cx] has joined #hplusroadmap
11:18 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has left #hplusroadmap []
12:12 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has joined #hplusroadmap
12:28 < docl> https://twitter.com/ESYudkowsky/status/1629210732612173825
13:11 < kanzure> https://openai.com/blog/planning-for-agi-and-beyond/ https://twitter.com/sama/status/1629212494072889349
13:12 < hprmbridge> lachlan> Seems like a reasonable way forward
13:15 < lsneff> except their note about open source
13:42 -!- o-90 [~o-90@gateway/tor-sasl/o-90] has joined #hplusroadmap
13:53 -!- o-90 [~o-90@gateway/tor-sasl/o-90] has quit [Ping timeout: 255 seconds]
14:00 -!- cthlolo [~lorogue@77.33.23.154] has quit [Read error: Connection reset by peer]
17:30 < hprmbridge> kanzure> https://www.modulardeeplearning.com/
19:02 < fenn> if you need a billion dollars to meaningfully compile the source code, it's still not really fair to open source things
19:04 < fenn> distributing pretrained models would bridge this gap somewhat, but still you're not going to be running it on a laptop
19:05 < fenn> i still think openAI is trying too hard to control things, for example chatGPT sucking more and more
19:08 < fenn> their policymaking process is very opaque
19:10 < kanzure> the in vitro cortical column stuff is pretty cool, maybe you can insert simulated blood vessels for nutrient distribution and then plug a bunch of these columns together on an alginate/hydrogel circuit board
19:11 < hprmbridge> kanzure> https://www.sciencedirect.com/science/article/pii/S2589004219303839
19:14 < fenn> brain tissue transplants ought to be fun and full of exciting plot twists
19:18 < muurkha> a tangled plot full of unexpected connections and neurotransmitters
19:25 < fenn> if i had severe brain damage and lost the ability to speak or make decisions, i'd want a brain tissue transplant
19:26 < fenn> just, you know, in case it comes up
19:26 < fenn> not that i'm very good a speaking or making decisions as it is...
19:26 < muurkha> noted
19:27 < muurkha> of course I don't have very good evidence that the entity known as "fenn" in this channel at this moment corresponds to any particular human body; it might just be a chatbot
19:27 < fenn> sure, and you might be a demon simulating all of "reality"
19:27 < muurkha> how did you guess!
19:28 < fenn> i didn't
19:28 < muurkha> you weren't supposed to figure it out for another 18 zlibniks
19:29 < muurkha> now we have to restart the simulation while we figure out what went wrong this time
19:29 < fenn> it won't matter; by refusing to take part in the free will experiment, my outputs are deterministic, and you'll just get the same results over and over
19:32 < fenn> i wonder if you bundled lots of these agarose/collagen hydrogel columns together, if the cells would branch out into neighboring columns in the same way as in cortical columns
19:37 < fenn> the problem with going all the way back to iPSCs instead of some more differentiated progenitor cell, is that you might not completely re-differentiate the iPSCs into neurons, and then get a teratoma. it would be bad to have teeth or hair growing in your brain
19:37 < muurkha> people often survive that way for decades
19:38 < muurkha> it might not be in the top thirty-two risk factors for your neuroscience experiment
19:39 < fenn> infection, epilepsy, what else?
19:40 < fenn> talked to a friend yesterday; he had a concussion and got some imaging done. they found lots of holes in his brain from an earlier stroke. he also has terminal cancer. seems like an ideal test subject
19:41 < fenn> but this being the world that it is, we won't see any brain tissue transplants until after we've all been turned into paperclips
19:41 < fenn> "too risky"
19:41 < fenn> "what if the family sues"
19:43 < muurkha> yeah, if I were your friend i would definitely be interested in being a test subject
19:44 < muurkha> some possible risks for implanting cell cultures to regenerate brains: tissue rejection, insufficient angiogenesis, psychosis, brain cancer that isn't a teratoma, necrosis for other poorly-understood reasons
19:45 < fenn> tissue rejection shouldn't happen because it's derived from the patient's cells, and the brain is an immune privileged region
19:46 < fenn> necrosis just means you're back where you started
19:46 < fenn> why would iPSC derived cells lead to cancer?
19:47 < hprmbridge> kanzure> are neurons really plug and play like that
19:49 < fenn> i'd like to find out!
19:49 < fenn> obviously the technique will require refinement, but also that paper is 4 years old
20:03 < hprmbridge> kanzure> "Interaction guided learning" https://arxiv.org/abs/2106.04887
20:10 < muurkha> necrosis hopefully means you're back where you started, yeah, but sometimes it creates vulnerability to infections
20:11 < muurkha> I think just about anything you do that makes cells more willing to proliferate increases the chance of cancer
20:11 < muurkha> you're surely right about the tissue rejection; I was somehow missing the context that this was the patient's own cells
20:12 < hprmbridge> kanzure> there was a different paper yesterday where it was xenotransplantation into visual cortex
20:50 -!- mrdata [~mrdata@user/mrdata] has joined #hplusroadmap
22:26 < fenn> more grist for the nightmare mill https://www.cdc.gov/prions/cwd/transmission.html
22:44 < muurkha> "CWD was transmitted to monkeys that were fed infected meat (muscle tissue) or brain tissue from CWD-infected deer and elk. Some of the meat came from asymptomatic deer that had CWD (i.e., deer that appeared healthy and had not begun to show signs of the illness yet)."
--- Log closed Sat Feb 25 00:00:54 2023