Damn this hits so hard

Brilliant

This guy is a machine elf in human form

Brilliant

not a very intelligent man interviewing

Seen you on the MLST channel also - fascinating stuff, thank you! (Tend to agree on politics these days, sheesh. We need to find ways to work together....and yes, focus on the important issues!!) Cheers...

stock images had me cracking up xD

Jeff Hawkins is asking questions about locations of mobiles here kzbin.info/www/bejne/ppCug6Z8YtOWgLM that I think could eventually be answered by the future GPT3… Here is my hypothesis: I was thinking of this last week : Brain makes maps of the environment. But I was wondering how it managed mobile objects. How it worked to know where it is. It's location. And then I remembered that when I was a kid, my mom would always ask me this question when I had lost something: When did you use it the last time? The question when. And I would try to remember what happened the last days, all the chronological events with a context linked to the object. And then I would remember where it could be. I'm amazed to consider that finally, the brain has a chronological map of events and their context, remembering every sequence and is able to navigate in this chronological map of events. And every object is there, the brain has all these objects identified and painted in every events, and keep all these pictures (I say picture, but it can be linked to any senses) chronologically and can navigate in this chronological mapping, like if the events were layers it can access. Soooooo if the events are like layers that can be accessed chronologically (or you can make jumps, but knowing where it stands in the chronology), could a future GPT3 do pattern recognitions on sequences. Could it learn sequences of patterns and then make predictions?

00:07:16.079,00:07:19.079 Yitz: howdy! 00:07:29.234,00:07:32.234 Yitz: doing well :) 00:07:32.117,00:07:35.117 Joshua Fox: Hello everyone 00:08:05.411,00:08:08.411 Yitz: this is my first meetup, looking forward to getting to know y'all! 00:08:41.536,00:08:44.536 Pythago Rascal: Hi all, anyone know how I set my name in this? 00:08:43.234,00:08:46.234 Roderick Ahrens: This is likewise my first online meetup. Looking forward to it! 00:09:51.078,00:09:54.078 Yitz: I assume I should be muted for now, right? 00:11:24.589,00:11:27.589 James Babcock: It mutes by default based on how many people are in the call 00:23:29.354,00:23:32.354 Isaac Leonard: www.gwern.net/Computers 00:25:35.339,00:25:38.339 Yitz: it is a pretty large assumption that nothing catastrophic happens, yeah :/ 00:26:39.843,00:26:42.843 Chlorkin: To clarify, one could imagine a future where instead of removing complexity, we use something like Codex to paper over it. 00:27:14.166,00:27:17.166 David Manheim: Systemic fragility to unexpected events doesn't get papered over well. 00:29:04.844,00:29:07.844 Yitz: what was important to you about the story of Braid? 00:29:44.620,00:29:47.620 Joshua Fox: Yitz, good question. You might want to ask in the Q&A 00:29:55.989,00:29:58.989 Yitz: asked there, sorry about that 00:30:05.713,00:30:08.713 Trevor Hill-Hand: I've been starting to think of game design as the art of showing a system through a carefully chosen lens. 00:33:00.624,00:33:03.624 Jonathan Blow: kzbin.info/www/bejne/hXi2ln-JorR0grc 00:33:21.615,00:33:24.615 Yitz: ooh, I've really been enjoying watching that game develop 00:33:28.546,00:33:31.546 Greg: I would like to play miegakure and I hope the developer has something playable soon 00:42:07.124,00:42:10.124 Yitz: Godel strikes again! 00:48:39.058,00:48:42.058 Yitz: he has said roughly as much in a recent blogg post 00:52:24.623,00:52:27.623 David Manheim: ...and the ways that specific laws get passed then circumvented tends not to eliminate the underlying pattern. 00:53:50.314,00:53:53.314 David Manheim: (The obvious way to fix most of this is for Google Play, Apples, etc. to limit the amount any one player can spend on a game's IAP. But that's against their interest in collecting a percentage, so it's probably not a real possibility.) 00:54:10.097,00:54:13.097 James Friesen: thanks for the answer jon! 00:57:17.494,00:57:20.494 Yitz: ethics-it's tricky! 01:05:19.367,01:05:22.367 Isaac Leonard: en.wikipedia.org/wiki/Beresheet 01:06:03.843,01:06:06.843 David Manheim: Isaac: What is that link supposed to relate to? 01:06:33.370,01:06:36.370 Isaac Leonard: Landing an archive on the moon. 01:06:49.651,01:06:52.651 Isaac Leonard: > The spacecraft carried a "time capsule" containing over 30 million pages of data, including a full copy of the English-language Wikipedia, the Wearable Rosetta disc,... 01:07:07.192,01:07:10.192 Isaac Leonard: (I think this is what he was referring to.) 01:07:52.817,01:07:55.817 Yitz: oh cool, my work on Wikipedia is on the Moon then! 01:10:28.767,01:10:31.767 Yitz: that's fine, as an artist myself I totally get it :) 01:11:11.803,01:11:14.803 Haru Glory: I love the longform answers :) 01:11:25.791,01:11:28.791 Trevor Hill-Hand: I see a lot of good questions, so my vote goes towards "as short as reasonable for the question, but no shorter" 01:12:46.766,01:12:49.766 Yitz: I think the current work of Nathalie Lawhead (creator of EVERYTHING IS GOING TO BE OKAY, etc.) is really intriguing in that respect 01:13:43.380,01:13:46.380 Yitz: en.wikipedia.org/wiki/Langton%27s_ant is my favorite cellular automata 01:15:12.170,01:15:15.170 mako yass: Jon should probably read the questions and decide for himself what would be interesting to answer, at least a few? He knows more than we do about what he has to say. 01:18:55.900,01:18:58.900 Yitz: would someone mind typing out the name of this book? I didn't catch the name 01:19:35.559,01:19:38.559 Mo Zhu: Baumol's cost disease 01:19:43.816,01:19:46.816 Yitz: thanks :) 01:19:53.067,01:19:56.067 Cameron Mulhern: I'm guessing it's this: www.goodreads.com/book/show/13594020-the-cost-disease 01:20:57.379,01:21:00.379 Caleb Rudnick: Thank you! 01:22:45.796,01:22:48.796 Greg: I think people are tempted to interpret Braid in that way because of the specific and rather jarring, way in which the atomic bomb reference happens in the game 01:22:59.372,01:23:02.372 Yitz: the best interpretations are those that leave you with even more questions than you had before, imo 01:25:36.084,01:25:39.084 Maxwell Joslyn: A good read about the technical aspects/challenges of decentralized systems: medium.com/swlh/a-unified-theory-of-decentralization-151d6f39e38 01:30:35.471,01:30:38.471 Yitz: I think you'd enjoy the altgame space; although much of the art in that scene isn't technically super impressive, there is an incredible amount of care and uniqueness put into it 01:32:44.765,01:32:47.765 Angelique Po: Thanks for answering. That's about what I was thinking re: AI-gen art.

I certainly don't want to distract you from what you are doing, but you made me fall in love! I even dreamed I was fixing your broken house's door so that you wouldn't have to take care of it and keep on working. Sooo important, go on, please

Re: 20:00 some comments mention - inadequate equilibria, simulacra, moral mazes, motive ambiguity, asymmetric justice, Copenhagen interpretation, blame avoidance, object-level models

Covid comments start at 16:15

this was a very useful presentation, espicially for people who do not use dmt to generate these states.

Chat from the meetup 00:03:47.044,00:03:50.044 David Manheim: Before hearing Bram's answer about Aumann Agreement, I want to note that I have a paper like 60% written that answers that question in detail, in the context of AI forecasting. 00:04:03.705,00:04:06.705 Joshua Fox: :-) 00:08:46.932,00:08:49.932 Joshua Fox: Hellow 00:09:08.426,00:09:11.426 David Manheim: Helloooooo! (Queen of England voice) 00:09:31.121,00:09:34.121 Vikki W.: I shared a Jam file with the meeting: jamboard.google.com/d/14R...[Redacted] 00:09:32.897,00:09:35.897 Vikki W.: I shared a Jam file with the meeting: jamboard.google.com/d/1BP...[Redacted] 00:10:29.068,00:10:32.068 David Manheim: I don't know what a jam file is, but it isn't shared with me... 00:11:33.394,00:11:36.394 Darij Grinberg: more like a Spartacus thing 00:12:04.925,00:12:07.925 Vikki W.: (The Jamboard might be world-readable now.) 00:13:59.507,00:14:02.507 David Manheim: Primecoin was a decent idea... 00:15:12.953,00:15:15.953 David Manheim: Are the proofs of space storing anything useful / specific? 00:15:28.507,00:15:31.507 Joshua Fox: Everyone, please ask questions and vote for questions. Click the triangle-square-circle icon in the lower right. Click "Q&A". 00:20:01.678,00:20:04.678 David Manheim: re: Marcin's q about the next big thing, it sounds like it's begging for a rationalist-fueled bubble. 00:22:45.937,00:22:48.937 David Manheim: Why doesn't proof of stake fulfill this goal? 00:22:53.305,00:22:56.305 Darij Grinberg: proof-of-availability <-- that will make DDOS really fun :) 00:26:27.019,00:26:30.019 Marcin Kowrygo: Could you specify it, David? To clarify, I kind of assume that a) similar opportunities are likely to appear in future, b) rational-ish circles might have advantage in detecting such opportunities the same way they had some advantage with Bitcoin/pandemic, c) opportunity window might be way shorter than the early days of cheap Bitcoin, so it might be important to stay alert. 00:33:28.735,00:33:31.735 David Manheim: Marcin: the problem is that the rationalist community now controls billions of dollars, easily. 00:34:00.977,00:34:03.977 David Manheim: So the window for noticing would need to be a subset of the community. 00:35:07.045,00:35:10.045 Joshua Fox: At the end we will meet and socialize in these rooms. I will explain more later sscmeetup.fastmai...[Redacted] 00:35:10.631,00:35:13.631 David Manheim: Interestingly, regarding the simultaneous nature of exchanges on Chia, this seems similar to batch-auctions instead of continuous clearing for markets. 00:41:49.662,00:41:52.662 David Manheim: As a user of Bitotrrent, the seed dropping off is a super-common issue, and would be fantastic to fix. 00:46:17.269,00:46:20.269 David Manheim: FEEDING PEOPLE was the biggest innovation in US education in the past ~30 years. 00:46:25.101,00:46:28.101 David Manheim: (i.e. school lunches.) 00:46:56.187,00:46:59.187 Joshua Fox: Innovation for the better? 00:47:05.870,00:47:08.870 Joshua Fox: I head that the kids just sneak off and buy junk food. 00:47:35.443,00:47:38.443 David Manheim: Joshua - kids rich enough to have money for junk food, sure. 00:47:51.519,00:47:54.519 Joshua Fox: OK, if that actually helps their nutrition, good 00:48:07.568,00:48:10.568 Garrett Baker: Kids don't need to sneak off & buy junk food. Often that is sold to them in school. 00:48:21.675,00:48:24.675 David Manheim: Joshua - test scores are a really simple proof - free lunches help a ton. 00:48:54.813,00:48:57.813 David Manheim: Start high schools after 10 AM. 00:49:16.945,00:49:19.945 David Manheim: Let the bus drivers sleep late! 00:50:11.705,00:50:14.705 David Manheim: And we're determined to make sure that kids need to get to school during rush hour, because otherwise traffic wouldn't suck as much. 00:50:13.880,00:50:16.880 Marcin Kowrygo: Ending up the wasteful credentialism/status signalling associated with formal education and (mostly) replacing the current system with homeschooling aligned with the 80,000 Hours content, boosted with MOOCs and online classes? 00:50:15.331,00:50:18.331 Vikki W.: I wonder what some "acceptable-enough-to-the-present-cultural-norms-it-would-even-be-considered" kind of argument that would promote more sleep for students / H.S. students would be. 00:51:00.997,00:51:03.997 David Manheim: For more on "let the kids sleep", follow @wendytroxel on twitter. 01:03:55.904,01:03:58.904 Garrett Baker: re: 4d visualization. Try 4dtoys.com/ 01:05:31.245,01:05:34.245 Vikki W.: I just asked my hubbie (who pays some attention to this stuff) what his estimate on number of ppl who can do 4D visualization, and he estimated 100-1000 ppl in the world atm. 01:06:42.922,01:06:45.922 Douglas Colkitt: It'd be interesting to find enough to do a GWAS on that subset. Or at least some correlation in terms of other factors 01:06:55.669,01:06:58.669 Garrett Baker: What's the basis of that estimate? 01:07:58.437,01:08:01.437 Vikki W.: Sorry, idk. I can get him to tell you sometime on a Saturday? (also, t.y. for that neat link!) 01:08:14.057,01:08:17.057 Marcin Kowrygo: We need to go deeper: www.5dchesswithmultiversetimetravel.com/ 01:08:28.450,01:08:31.450 Garrett Baker: @vikki Cool. You're welcome 01:11:08.611,01:11:11.611 Chlorkin: "Modal human" 01:12:10.617,01:12:13.617 Jorge Azpurua: I'm a molecular biologist by trade. Everything said so far is accurate, in case anyone wanted a spot check. 01:13:03.991,01:13:06.991 Douglas Colkitt: UBI weighted by mutational load 01:14:12.015,01:14:15.015 David Manheim: Spreading rapidly in modern humans, across populations, now, is probably not a good feature... 01:14:57.037,01:15:00.037 Jorge Azpurua: Sometimes things can spread rapidly due to linkage disequilibrium during selective sweeps. Or because they are segregation distorters. 01:16:47.893,01:16:50.893 Jorge Azpurua: *applause* 01:16:53.220,01:16:56.220 Joshua Fox: At the end we will meet and socialize in these rooms. I will explain more sscmeetup.fastmai...[Redacted] (and yes, those room names are SHA256 hashes). 01:16:55.675,01:16:58.675 Marcin Kowrygo: *counterfactual clapping* 01:16:57.272,01:17:00.272 David Manheim: For most of these gross phenotypical features, we also have no idea what other things they cause...

Well, I know this guy as the leader of ElutherAI, so I'd say he's passed that benchmark.

Edited chat transcript ------------------ 00:32:38.517,00:32:41.517 Luke: Surprised he didn't say cryptography as first application area 00:32:45.599,00:32:48.599 David Manheim: Maybe it's just me, but I keep mishearing "q-bits" as "cubits" 00:33:19.467,00:33:22.467 John Kaniarz: Noha's Ark as a quantum device. 00:37:51.220,00:37:54.220 Shiv Shankar: that might also test one of penrose's idea of gravity connected with collapse 00:38:11.261,00:38:14.261 David Manheim: Wondering what von-Neumann thought computers would be useful for... 00:38:38.864,00:38:41.864 Patrick: Weather modification... 00:39:57.429,00:40:00.429 David Manheim: I have a forthcoming (being drafted) paper on AI elicitation which discusses why we shouldn't expect it to hold. 00:40:50.639,00:40:53.639 Mikhail Samin: Though there's the street epistemology thing a lot of people are doing IRL 00:41:01.901,00:41:04.901 David Manheim: Just hearing their opinion would be sufficient... if computation was costless. 00:41:02.909,00:41:05.909 Shiv Shankar: is this like the ideal free market idea, where no trade actually happens? 00:41:26.642,00:41:29.642 Patrick: @Samin Socrates? 00:41:38.748,00:41:41.748 Shiv Shankar: ideal free market for stocks and contracts* 00:42:36.684,00:42:39.684 David Manheim: Play the game and you can watch it happen - www.lesswrong.com/posts/nmwog5hGidZniDDpR/aumann-agreement-game 00:46:11.267,00:46:14.267 Marcus Wilhelm: Scott was a teenager when what was discovered? my connection stumbled on that one 00:46:19.846,00:46:22.846 Michael Zhang: It's not so much we're stagnating now as technological change happened at exceptional pace for the first century after the Industrial Revolution 00:46:53.643,00:46:56.643 Luke: arxiv.org/pdf/2001.04383.pdf 00:47:03.768,00:47:06.768 PARAKRAM SINGH SHEKHAWAT: thanks 00:47:38.408,00:47:41.408 Deborah Art-Weisman: Verner Vinge did! 00:50:15.324,00:50:18.324 Marcus Wilhelm: isn't it fascinating how research in maths still continues after thousands of years? 00:50:28.267,00:50:31.267 Luke: So far to go 00:51:31.890,00:51:34.890 Michael Zhang: don't worry, it's the same in my field 00:51:36.899,00:51:39.899 David Manheim: It might be worse in Quantum Computing, but the failure of journalism is pretty broad. 00:53:45.517,00:53:48.517 Mr R: What do you think about Nielsen and Chuang? 00:53:47.632,00:53:50.632 Shiv Shankar: vazirani notes are amazing... 01:02:05.823,01:02:08.823 Shiv Shankar: motivating reviewers to do get good reviews is hard. 01:02:06.861,01:02:09.861 Luke: openreview.net has been quite successful 01:02:16.298,01:02:19.298 Luke: I think they just publish the reviews and call it quits 01:02:33.890,01:02:36.890 Luke: Seems like a pretty good solution to me 01:03:47.822,01:03:50.822 Yevgeny Liokumovich: Bourbaki? :-) 01:04:50.108,01:04:53.108 Braxton Hartman: eLife does open reviews, and has been vey successful in life sciences 01:05:08.394,01:05:11.394 David Manheim: If I had invested in Bitcoin the first time I heard about it... I'd have lost hundreds of millions of dollars on a hard drive I threw out a couple years later after forgetting about it 01:06:48.065,01:06:51.065 Joseph Kamien: I need to drop off. Thank you everyone for a great session! 01:12:26.400,01:12:29.400 David Manheim: Fight! 01:12:35.540,01:12:38.540 PARAKRAM SINGH SHEKHAWAT: fight 01:12:38.073,01:12:41.073 PARAKRAM SINGH SHEKHAWAT: fight 01:12:39.438,01:12:42.438 PARAKRAM SINGH SHEKHAWAT: fight 01:12:46.580,01:12:49.580 Ariel Krakowski: matzah with cheese 01:12:54.487,01:12:57.487 Gregory Rosenthal: Bitter herbs are horrible, other Passover food is good. 01:13:50.795,01:13:53.795 Stephen Boyd: Oh nvm you already did. 01:17:13.072,01:17:16.072 David Manheim: (Now is when I get to brag that Anders convinced Scott to give us feedback on our limits to value paper.) 01:24:55.908,01:24:58.908 Michael Zhang: seems typical of crackpots in all fields 01:26:55.100,01:26:58.100 David Manheim: No infinities! 01:27:25.557,01:27:28.557 Shiv Shankar: @david cantor wants to have a word with you :D 01:28:08.458,01:28:11.458 Michael Zhang: I'll book a room for both of you in Hilbert's Hotel 01:29:12.386,01:29:15.386 Alex Sato: 42 01:29:54.096,01:29:57.096 Patrick: Adamsian ultrafinitism 01:31:06.150,01:31:09.150 Luke M: www.scottaaronson.com/blog/?p=1753 01:31:19.193,01:31:22.193 Shiv Shankar: thanks luke 01:31:23.128,01:31:26.128 Stephen Boyd: Thank you! 01:35:08.708,01:35:11.708 Shiv Shankar: From one of scott's comments, does any one have any information about the decidable universe concept he talked. I am pretty sure I have seen work which show purely analog computation can be super-turing 01:48:50.493,01:48:53.493 Shiv Shankar: this is russell's paradox question :D 01:49:13.032,01:49:16.032 Steven: "Nobody asked me what my favorite food was!"

Lol at the twitter feed question

Is there any good material on the overlap of Qualia Research and deep meditation practice? There seems so much parallels for me as somehow who has dangled in both. For the topic of heaven realms for example the jhanic concentration states come to my mind which are beyond blissful and also called the heavenly abodes.

Awesome. Great presentation!

What is QRI's view on free will?

Does the example of extremely high positive valence in the experience of becoming a parent somehow contradict the whole 'replicators vs consciousness' notion?

1:14:00 "doobadoop yeah" great presentation

Edited chat transcript docs.google.com/document/d/1FOGshVwwxNhGYCqoSyy8cPij_oFsmYwWF4jW1QV4QLU/edit

Major timestamps up to 101:50: 00:00 introductory remarks 27:30 Q: ..Should we generalize rationalist confidence? Is it good? 31:30 We realized the emperor had no clothes! 34:01 Q: What blockchain will the game run on...? 35:59 Q: What are your disagreements with mainstream scientific consensus on COVID-19 and pandemic management? 40:15 Q: why did America and EU countries get things so bad? 43:20 Q: blockchains vs centralized marketplaces 47:40 Q: Did SARS-CoV-2 originate in a lab? 52:50 "Bioethicists are to blame for entirety of the COVID crisis..." 53:00 Q: How well do the planned economics of a card game hold up? 58:30 Q: Will the set of cards available be known beforehand? 1:01:45 Q: What did Zvi get right, what did he get wrong? (ie is survivership bias in predictions an issue here?)

Chat transcript here docs.google.com/document/d/1t2TAspEes3gw6Bo5KLs8dMrBdGL5QumgFLi02CfmtMo/edit?usp=sharing

Actual start is at 4:48

You mispelled plandemic

Talk begins at 2:50

Dude slow down - can click that like button only once..

Thanks for the presentation Connor. Most intriguing.

Starts at 13:30

Good talk, GPT-3 should be the fire alarm that wakes everyone up. AI safety research is the main field that is going to make a change long term.

Considering the ant analogy w.r.t. the hydro dam: even if we could explain it in ant-language, the fact that they have to move their whole society means they probably just won't do a damn thing: see climate change. also: Good example of the fire alarm, never thought about it that way.

This link in the description does not work: "Video on GPT-3: Language Models are Few-Shot Learners"

You mentioned you have experimented with GPT-3. I would love to do so myself, but I was not able to find GPT-3's model and weights available. Is it only available for certain people? Or you have to register somewhere?

27:54 Not only is our planet not new, but it is thought that it took well over a billion years after life started before it got lucky and created the innovation of sex, and the specialization provided by mitochondria, and then more than another billion years for multicellular Eukyrotic life to evolve. Presumably our luck could have been better (but it could equally be that our timeline was unusually luckily.) 31:56 If I lived to be a trillion years old, I'd probably decide that it is likely I will want to live to be a quadrillion years old and therefore I'd want to start capturing the energy of stars (and do what with it? A better approach might be to somehow prevent fusion from starting inside nebulas until it is deliberately triggered, thus creating "battery stars" to be activated trillions of years later). But in a younger universe like ours, perhaps there is little urge to actually do this. To the extent there is an urge, Dyson spheres could be hard to notice once completed, and our tools of observation are limited. 42:40 All evidence points to the simulation hypothesis being false. If it is true, the simulation must devote a lot of processing power to deceiving us with evidence that it is false, and therefore the simulation was designed for us specifically. But then, why deceive us with evidence and yet allow people to believe in it anyway? medium.com/big-picture/sorry-youre-not-in-the-matrix-4a321eed8384 cosmosmagazine.com/physics/physicists-find-we-re-not-living-in-a-computer-simulation

17:05 The formation of intelligent life could be extremely unlikely, putting a Great Filter behind us, but that doesn't mean there isn't another one in front of us. Keep this in mind before deciding that a Paperclip Maximizer (or happy face maximizer, or happy-acting otherwise-useless hominid maximizer) is unlikely. Or, if civilizations rarely make them because it's such an obvious trap, we must still be careful to avoid the trap. Also note that an organization that makes a dangerous Paperclip Maximizer necessarily does so by accident, and so may tend not design it in such a way that it expands beyond its home planet, or it may eventually become self-defeating, which would explain why we wouldn't notice it. Another version of great filter is where any biological civilization that expands will eventually colonize a planet ruled by a paranoid dictator who decides that all other colonies are a potential threat and decides to kill them all off with biological weapons, sent on probes, before they can attack. Synthetic civilizations could face an analogous problem in a form that is unpredictable (if they are intelligent enough, they might develop decision theory to the point where they detect that expansion is inherently unsafe and must not be done; of course it is not possible to know whether this scenario is plausible without ourselves being as advanced as they are). 24:40 Err... huh? what? See Stargate Universe Season 2 (fourth quarter) for a depressing scenario that could be imagined in which it would be extremely unwise to broadcast our presence to the rest of the galaxy. In all cases we must remember that destruction is generally easier than creation, in our universe anyhow. Note also that in recent years we ourselves have reduced our output of intelligent-looking radio signals (digital signals, as they become more efficient, tend to both get weaker, and more focused/directed, and resemble noise to a greater and greater extent), implying it is hard to find signals from advanced civilizations unless they broadcast intentionally.