--- Log opened Mon Jan 15 00:00:29 2024
04:06 -!- darsie [~darsie@84-113-55-200.cable.dynamic.surfer.at] has joined #hplusroadmap
04:55 -!- faceface [~faceface@user/faceface] has joined #hplusroadmap
05:27 -!- ike8 [e8f913dbdf@irc.cheogram.com] has joined #hplusroadmap
05:59 -!- mxz [~mxz@user/mxz] has joined #hplusroadmap
06:02 -!- mrdata_ [~mrdata@135-23-182-55.cpe.pppoe.ca] has joined #hplusroadmap
06:05 -!- mrdata [~mrdata@user/mrdata] has quit [Ping timeout: 256 seconds]
06:31 < fenn> naming is hard: https://singularityhub.com/2023/12/14/a-ball-of-brain-cells-on-a-chip-can-learn-simple-speech-recognition-and-math/
06:46 -!- flooded [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
06:47 -!- TMM_ [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
06:47 -!- TMM_ [hp@amanda.tmm.cx] has joined #hplusroadmap
08:40 -!- mrdata_ [~mrdata@user/mrdata] has changed host
08:55 -!- test__ [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
08:59 -!- flooded [flooded@gateway/vpn/protonvpn/flood/x-43489060] has quit [Ping timeout: 245 seconds]
09:43 -!- mrdata_ is now known as mrdata
09:46 < hprmbridge> Eli> Silicon computing vs carbon computing 🤔
09:47 < hprmbridge> kanzure> we should grow superintelligent animals and use them as AGI agents.
09:54 < hprmbridge> Eli> Is neuralink going to be the first AGI agent? 🤔
09:56 < L29Ah> no nose picking emoticon in unicode :/
09:57 < hprmbridge> Eli> You can submit one. There is plenty of space and new emoticons are approved all the time
09:59 < hprmbridge> Eli> Tell us more
09:59  * superkuh imagines a FUSE filesystem that stores data by submitting tiny qr codes to the unicode spec for approval/inclusion.
10:41 -!- Gooberpatrol66 [~Gooberpat@user/gooberpatrol66] has joined #hplusroadmap
10:47 < hprmbridge> kanzure> Neuralink is just an implant, it needs silicon for the offloading. Not the same thing.
10:48 < hprmbridge> kanzure> soon. are you coming back for april?
10:48 < hprmbridge> kanzure> same question to superkuh
10:49 < superkuh> If this is re: meet-up, no. I am still not traveling until I can get intranasal vaccination for sars-cov-2.
11:06 < hprmbridge> kanzure> hmph
11:27 -!- flooded [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
11:31 -!- test__ [flooded@gateway/vpn/protonvpn/flood/x-43489060] has quit [Ping timeout: 264 seconds]
12:03 < L29Ah> 👃 ⛏️
12:26 < hprmbridge> Eli> Would love to.
13:38 -!- Gooberpatrol66 [~Gooberpat@user/gooberpatrol66] has quit [Ping timeout: 256 seconds]
13:59 -!- Gooberpatrol66 [~Gooberpat@user/gooberpatrol66] has joined #hplusroadmap
14:04 -!- Gooberpatrol66 [~Gooberpat@user/gooberpatrol66] has quit [Ping timeout: 260 seconds]
14:18 -!- Gooberpatrol66 [~Gooberpat@user/gooberpatrol66] has joined #hplusroadmap
14:20 -!- test__ [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
14:22 -!- flooded [flooded@gateway/vpn/protonvpn/flood/x-43489060] has quit [Ping timeout: 260 seconds]
14:32 -!- Gooberpatrol66 [~Gooberpat@user/gooberpatrol66] has quit [Ping timeout: 240 seconds]
14:48 -!- Lando-HoloNet [~Lando-Spa@user/lando-spacepimp] has joined #hplusroadmap
14:50 -!- Hooloovoo [~Hooloovoo@hax0rbana.org] has quit [Remote host closed the connection]
14:51 -!- Ashstar [~Ashstar@mobile-166-170-41-40.mycingular.net] has quit [Ping timeout: 264 seconds]
14:52 -!- Lando-SpacePimp [~Lando-Spa@user/lando-spacepimp] has quit [Ping timeout: 268 seconds]
14:53 -!- Hooloovoo [~Hooloovoo@hax0rbana.org] has joined #hplusroadmap
14:53 -!- Ashstar [~Ashstar@mobile-166-170-41-40.mycingular.net] has joined #hplusroadmap
15:03 -!- Netsplit *.net <-> *.split quits: L29Ah, superkuh, SDr
15:04 -!- Netsplit over, joins: L29Ah, superkuh, SDr
15:13 -!- Gooberpatrol66 [~Gooberpat@user/gooberpatrol66] has joined #hplusroadmap
15:38 -!- darsie [~darsie@84-113-55-200.cable.dynamic.surfer.at] has quit [Ping timeout: 255 seconds]
16:30 -!- flooded [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
16:34 -!- test__ [flooded@gateway/vpn/protonvpn/flood/x-43489060] has quit [Ping timeout: 256 seconds]
18:09 -!- faceface [~faceface@user/faceface] has quit [Ping timeout: 264 seconds]
18:10 -!- faceface [~faceface@user/faceface] has joined #hplusroadmap
18:21 -!- stipa [~stipa@user/stipa] has quit [Ping timeout: 260 seconds]
18:22 -!- stipa [~stipa@user/stipa] has joined #hplusroadmap
18:29 -!- TMM_ [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
18:29 -!- TMM_ [hp@amanda.tmm.cx] has joined #hplusroadmap
18:46 < hprmbridge> kanzure> mergekit/MoE stuff https://twitter.com/main_horse/status/1746853576184668443
19:02 -!- test__ [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
19:05 -!- flooded [flooded@gateway/vpn/protonvpn/flood/x-43489060] has quit [Ping timeout: 246 seconds]
19:53 -!- Gooberpatrol66 [~Gooberpat@user/gooberpatrol66] has quit [Quit: Leaving]
20:00 -!- Gooberpatrol66 [~Gooberpat@user/gooberpatrol66] has joined #hplusroadmap
20:50 -!- mxz [~mxz@user/mxz] has quit [Ping timeout: 256 seconds]
21:33 -!- flooded [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
21:35 < fenn> the reason you can ask an LLM what its strengths are is that they know what they do and do not know. this is a good paper about it: https://arxiv.org/abs/2207.05221
21:35 < fenn> .t
21:35 < EmmyNoether> [2207.05221] Language Models (Mostly) Know What They Know
21:36 < fenn> re "it sounds counterintuitive but the most effective solution is to load the individual ai models that you are using for your merge and ask them what their strengths are."
21:36 < fenn> it won't work for logic and arithmetic capabilities
21:37 -!- test__ [flooded@gateway/vpn/protonvpn/flood/x-43489060] has quit [Ping timeout: 252 seconds]
21:39 < muurkha> hey, that puts them way ahead of the humans
21:39 < fenn> .t https://arxiv.org/abs/2305.14975
21:39 < EmmyNoether> [2305.14975] Just Ask for Calibration: Strategies for Eliciting Calibrated Confidence Scores from Language Models Fine-Tuned with Human Feedback
21:40 < fenn> this paper discusses ways to get the info. it turns out asking them in natural language works better than other more technical ways. "Verbalized confidence scores are better-calibrated than log probabilities for gpt-3.5-turbo"
21:42 < fenn> muurkha: humans have evolved social status improving behaviors to make it look like they're smart even if they aren't
21:42 < fenn> LLMs haven't had the opportunity to evolve in a social environment
21:43 < fenn> maybe we are seeing an analogous process with the benchmark leaderboard cheating and benchmark leakage into training data
21:46 < fenn> the nose pick ligature would be readable and backwards compatible
21:46 < fenn> but it's supporting the proliferation of emoji culture, which i cannot support
21:46 < fenn> death to all emoji
22:01 < fenn> .t https://1a3orn.com/sub/machine-learning-bans.html
22:01 < EmmyNoether> Many AI Safety Orgs Have Tried to Criminalize Currently-Existing Open-Source AI
22:18 < muurkha> at least they aren't assassinating its users yet
22:19 < fenn> it looks like they copy-and-pasted the executive summary from the center for AI safety's "AI regulatory authority" into the document used to create biden's executive order
22:21 < fenn> i've never even heard of the National Telecommunications and Information Administration (NTIA) but it seems like the wrong agency to be in charge of this
22:22 < fenn> "NTIA’s programs and policymaking focus largely on expanding broadband Internet access and adoption in America, expanding the use of spectrum by all users, and ensuring that the Internet remains an engine for continued innovation and economic growth."
22:36 -!- frege [uid547359@id-547359.hampstead.irccloud.com] has joined #hplusroadmap
22:39 < fenn> as if the regular "Code of Conduct" (mandatory) weren't bad enough, now there's a G7 International Code of Conduct
22:40 < fenn> so that developers can't escape from overbearing regulation and outcompete those stuck in the over-regulated states
22:43 < hprmbridge> kanzure> aren't we supposed to route around this stuff or something
22:45 < muurkha> that was Usenet
22:46 < hprmbridge> kanzure> ew, old software? no thanks.
22:47 < muurkha> it's not so much the software as the social structure, I think
22:47 < fenn> google has banned usenet, fyi
22:48 < muurkha> yeah
22:48 < fenn> https://www.pcmag.com/news/end-of-an-era-google-groups-to-drop-usenet-support
22:50 < fenn> " If anything, having Google drop off of Usenet is going to be a benefit for the net because it will eliminate the spam without actually losing many legitimate users.  So many backbone sites already drop Google postings because of their spam that there aren't many legitimate Google usenet users left anyway."
22:50 < Hooloovoo> who uses usenet nowadays?
22:51 < Hooloovoo> I'm in old computer channels and don't know anyone who's got their vax or whatever hooked up
22:51 < Hooloovoo> or anyone who uses it for normal stuff, even
22:53 < muurkha> it's not widely used these days
22:53 < muurkha> but Usenet was the net that treats censorship as damage and routes around it
22:53 < muurkha> not the internet
22:54 < fenn> i can't help but read "ew, old software?" in a valley girl voice
22:55 < fenn> "Atomkraft? Nein danke"
22:56 < muurkha> also though Gilmore's remark predated Dick DePew and ARMM's robocancels
22:56 -!- mxz [~mxz@user/mxz] has joined #hplusroadmap
22:57 < muurkha> when you build social software you are inscribing your social discourse into the computer.  the results depend on which social discourse that is
23:00 < muurkha> computers are universal; they will implement whatever dynamics you program them to implement
23:04 < muurkha> I mean, there are things that are not so flexible, like in crypto.  But that isn't what happened to Usenet, I think.
23:11 < fenn> "content authentication and provenance mechanisms" sounds especially dreadful
23:12 < fenn> it sounds like they are gearing up to require watermarking on everything
23:12 < fenn> (nevermind that nobody knows how to do that)
23:14 < fenn> why not just ban AI? it would be simpler and more effective
23:14 < fenn> it would have been impossible to achieve the current state of AI capabilities if these regulations had been in place already
23:18 < muurkha> well, if you think the current state of AI capabilities is already too powerful, that's probably a recommendation
23:20 < fenn> well, i don't
23:21 < fenn> we're in a mushy middle ground where we get all the "harms" and none of the benefits
23:22 < Ashstar> hey
23:22 < fenn> legislating software has never worked out well
23:22 < Ashstar> sometimes, mushy is good
23:23 < Ashstar> or not
23:24 -!- darsie [~darsie@84-113-55-200.cable.dynamic.surfer.at] has joined #hplusroadmap
23:26 < Ashstar> I remember some of the studies that one of my former advisors did
23:27 < Ashstar> dealing with voluntary hypometabolism 
23:28 < Ashstar> there seems to be a state that can be achieved where the aerobic metabolism is reduced to >2%
23:29 < Ashstar> without lactogenosis
23:31 < Ashstar> https://www.sciencedirect.com/science/article/abs/pii/030645658790060X
23:34 < Ashstar> it requires a focused metitative mind
23:34 < Ashstar> meditative
23:36 < Ashstar> I belive
23:39 < Ashstar> on another note: https://www.scopus.com/record/display.uri?eid=2-s2.0-85127327466&origin=inward&txGid=4657375cae9a17c4eff3b149fa0907a1
23:43 < fenn> link is broken
23:45 < Ashstar> there was a time I got really serious about lifestyle, health, and meditation
23:45 < Ashstar> I had a friend who had inspired many
23:46 < Ashstar> lived near me, I went to his place, he was in this deep state of meditation, I left
23:46 < Ashstar> came back several hours later
23:47 < Ashstar> he was like in this state, still, breathless
23:47 < Ashstar> it was fascinating to see him, suspended
23:48 < Ashstar> in something that I felt was really special
23:48 < Ashstar> it kicked my ass to go there
23:49 < Ashstar> I spent time trying to reach that state
23:50 < Ashstar> then I felt it, the suspended focused state of mind
23:50 < muurkha> how's your blood oxygenation?
23:50 < Ashstar> loss of time
23:50 < Ashstar> no breath
23:50 < Ashstar> it was a sort of immersion
23:51 < Ashstar> to stateless state of time
23:51 < Ashstar> then, my body reminded me
23:51 < Ashstar> I wanted to go back
23:52 < Ashstar> I left that experience feeling there was something that time, death could not touch
23:53 < Ashstar> my ability to focus was so much more capable
23:54 < Ashstar> that was an experience, hard to describe
23:54 < Ashstar> as far as O2
23:55 < Ashstar> I thought there was a switch
23:55 < Ashstar> to an anaerobic state
23:56 < Ashstar> by reducing metabolism to less than 2%
23:58 < Ashstar> I was determined to find the science, if any, behind this
23:59 < Ashstar> just hope to recapture it before I exit
23:59 < Ashstar> for myself
23:59 < Ashstar> I think there is a metabolic switch
--- Log closed Tue Jan 16 00:00:31 2024