--- Log opened Tue Apr 04 00:00:22 2023
01:02 -!- cthlolo [~lorogue@77.33.23.154.dhcp.fibianet.dk] has joined #hplusroadmap
02:16 < fenn> "The majority of the 738 researchers who responded to a recent survey (targeting those who published recently at machine-learning venues NeurIPS and ICML) assigned a greater than 10% chance of 'human inability to control future advanced AI systems causing human extinction'" (Stein-Perlman et al. 2020)
02:18 < fenn> 2022 survey https://aiimpacts.org/2022-expert-survey-on-progress-in-ai/#Extinction_from_human_failure_to_control_AI
02:20 < fenn> i guess what i'd really like to know is how many of the respondents were typical academic researchers vs how many are yuddites and responded to the survey because it was distributed in yuddite circles
02:23 < fenn> i think the 2020 citation was an error and it was actually the 2022 survey
02:33 < fenn> over 99% responded "Academia" or "Industry" and less than 1% responded "Other"
03:32 -!- flooded [~flooded@146.70.202.142] has joined #hplusroadmap
03:35 -!- test_ [~flooded@146.70.174.179] has quit [Ping timeout: 268 seconds]
05:27 -!- Llamamoe [~Llamamoe@46.204.73.97.nat.umts.dynamic.t-mobile.pl] has joined #hplusroadmap
05:50 -!- yashgaroth [~ffffffff@2601:5c4:c780:6aa0:92a:5964:b242:7adb] has joined #hplusroadmap
06:41 -!- test_ [~flooded@146.70.174.227] has joined #hplusroadmap
06:45 -!- flooded [~flooded@146.70.202.142] has quit [Ping timeout: 260 seconds]
06:55 -!- test_ is now known as _flood
07:24 < hprmbridge> kanzure> https://youtu.be/MyNHpMoPkIg
07:24 < Muaddib> [MyNHpMoPkIg] Draft Ethical Principles of Therapeutic Assisted Reproductive Technologies (5:15)
07:24 < hprmbridge> kanzure> did we ever archive the JK videos ?
07:44 -!- cthlolo [~lorogue@77.33.23.154.dhcp.fibianet.dk] has quit [Ping timeout: 250 seconds]
08:46 -!- cthlolo [~lorogue@77.33.23.154.dhcp.fibianet.dk] has joined #hplusroadmap
12:15 < L29Ah> https://www.lifespan.io/news/bryan-johnsons-race-against-time/ the lab rat humanity needed
12:59 -!- flooded [~flooded@169.150.254.33] has joined #hplusroadmap
13:02 < fenn> iirc the JK videos were taken down within a day or so of the original announcement, and i for one didn't archive them
13:04 -!- _flood [~flooded@146.70.174.227] has quit [Ping timeout: 276 seconds]
13:05 -!- cthlolo [~lorogue@77.33.23.154.dhcp.fibianet.dk] has quit [Read error: Connection reset by peer]
13:05 < fenn> "Johnson wakes up naturally between 4:30 and 5:30 am" --> already winning
13:08 < fenn> hmm nevermind, he goes to bed at 8:30 pm
13:09 < hprmbridge> lachlan> 8:30 is absurdly early to go to bed wow
13:10 < L29Ah> the time of day doesn't matter; what matters is stability
13:11 < fenn> i'm sure the light pollution of modern environments does matter in the end
13:11 < L29Ah> if you have a room, having a dark room is cheap
13:12 < fenn> i mean, people using light bulbs after sunset, computer screens, etc
13:12 < fenn> having it be dark indoors when it's dark outdoors, not only when you're sleeping
13:13 < fenn> perhaps a better thing to try to replicate is bryan johnson's insane level of dedication
13:14 < fenn> like yeah, these things all sound like probably a good idea, but i haven't done any of them
13:14 < fenn> in the end we're all made out of meat, so what is actually the difference between me and him?
13:15 < L29Ah> it doesn't look like he has lots of dedication tbh
13:15 < L29Ah> probably the hardest part is the CR thing, keeping body fat at 5%
13:24 < fenn> if i were johnson i'd enroll an army of borg drones to help resolve ongoing questions about the efficacy of specific interventions
13:25 < fenn> all completely voluntary of course
13:25 < fenn> well, dying of old age is involuntary, so we're all forced into it i suppose
13:30 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has quit [Quit: Gateway shutdown]
13:39 < fenn> is micromorts really a useful concept? is doing nothing for one additional day really as likely to kill you as skydiving 3 times?
14:04 < fenn> there has to be some way to reformulate this so that it makes sense
14:04 < muurkha> I don't know, what fraction of regular skydivers die skydiving?
14:05 < muurkha> I feel like micromorts is really a useful concept but maybe there's a problem in your calculation there
14:06 < muurkha> like maybe you're taking the average of doing nothing for one additional day over all age groups, rather than just a particular age group
14:06 < muurkha> when you're 102, doing nothing for one additional day is pretty hazardous
14:09 < fenn> what's the overall difference in utility if you waste a day at age 10 vs age 100
14:11 < fenn> at age 10 everything is unique and special and amazing, and at 100 you're probably pretty jaded and stuck in a routine
14:11 < kanzure> it is unclear if the whole "getting old and stuck with old thoughts/routine" is a real problem or just a choice that people make (or maybe it's mostly biological)
14:12 < kanzure> sunk cost fallacy as far as the eyes can('t) see
14:12 < fenn> well.. if we had radically legal psychedelics research it might help answer that question
14:13 < fenn> also at the risk of sounding like a broken record, lion's mane mushroom neurogenesis feels like turning into a different person
14:13 < fenn> i don't know if that's "good" or "bad"
14:15 < kanzure> the videos look online to me: https://www.youtube.com/@thehelab4617
14:16 < kanzure> one of them has 688 kiloviews but the others <82 kiloviews
14:16 < kanzure> 688 kiloviews is kind of a low number given the amount of time people spent complaining about it
14:23 < yashgaroth> tbh that's higher than I thought it'd be
14:28 < kanzure> dearluluandnana@gmail.com
14:28 < kanzure> didn't know about this address... maybe a few wise words are in order.
14:28 < fenn> what's it for?
14:29 < fenn> honeypot for anti-gmo nutters?
14:31 < kanzure> i think they are letters that will be available to the children in the future
14:50 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has joined #hplusroadmap
15:09 -!- Llamamoe [~Llamamoe@46.204.73.97.nat.umts.dynamic.t-mobile.pl] has quit [Quit: Leaving.]
15:11 < kanzure> apparently people are using it to just rant:
15:11 < kanzure> https://artcompass.medium.com/my-letter-to-lulu-and-nana-the-worlds-first-crispr-gene-edited-babies-77aa4ecec2fd
15:12 < kanzure> or this weird one:
15:12 < kanzure> https://www.makery.info/en/2021/09/01/english-human-germline-gene-editing-is-bioart-an-open-letter-to-lulu-and-nana/
15:12 < kanzure> where the author compares his reader to... glofish?
15:14 < kanzure> how about a letter that just tells them "yo don't go public"
16:40 -!- Gooberpatrol66 [~Gooberpat@user/gooberpatrol66] has quit [Remote host closed the connection]
16:44 -!- Jay_Dugger [~jwd@47-185-225-108.dlls.tx.frontiernet.net] has joined #hplusroadmap
17:10 < fenn> i've done nothing today but go squee about starship and pace nervously
17:11 < kanzure> maybe we should rent a van
17:11 < kanzure> or a bus
17:12 < fenn> are you interested in going?
17:13 < kanzure> well.... the house might have people.. but maybe a bus for everyone else.
18:24 -!- Jay_Dugger [~jwd@47-185-225-108.dlls.tx.frontiernet.net] has quit [Ping timeout: 268 seconds]
18:40 < muurkha> kanzure: it seems likely that some part of the shift from exploration to exploitation that is optimal in all reinforcement learning games is biologically programmed
18:41 < kanzure> hm? you can't do both in different domains at the same time?
18:41 < muurkha> you can do both in the same domain at the same time
18:42 < muurkha> but as you have more experience to work from, it's valuable to do more exploitation and less exploration
18:50 -!- yashgaroth [~ffffffff@2601:5c4:c780:6aa0:92a:5964:b242:7adb] has quit [Quit: Leaving]
19:18 < fenn> there wasn't as much change in the ancestral environment, maybe
19:18 -!- test_ [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
19:18 < fenn> so there would be less value to exploration late in life
19:20 < muurkha> there still is less value to exploration late in life than early in life, though more than in the ancestral environment
19:21 -!- flooded [~flooded@169.150.254.33] has quit [Ping timeout: 248 seconds]
19:21 < muurkha> I mean, if you learn Spanish when you're 50, you can speak Spanish for 30 years.  but if you learn Spanish when you're 5, you can speak Spanish for 70 years.  So you get twice as much benefit from the same exploration cost
23:19 < fenn> HuggingGPT lets ChatGPT pick other models on huggingface to fulfill sub-tasks https://github.com/microsoft/JARVIS#overview
23:21 < fenn> i wonder when openAI will provide LoRA creation/sharing capability
23:22 < fenn> since LLMs with memory are turing complete, it might have some nasty security problems
23:57 -!- cthlolo [~lorogue@77.33.23.154.dhcp.fibianet.dk] has joined #hplusroadmap
--- Log closed Wed Apr 05 00:00:23 2023