Robin Hanson - The Long View and The Elephant in the Brain
Skills:
AI Alignment Basics80%
Key Takeaways
Examines the long view and the elephant in the brain with Robin Hanson, professor of economics at George Mason University
Full Transcript
[Music] okay robin hansen needs our introduction so let's begin so you've written a few essays on uh the longview how there's no second flair significant players that are optimizing for it but isn't the most obvious explanation for that that uh optimizing for the long term is ineffectual given how unpredictable the future is or even counterproductive given uh that you might be ignoring present circumstances well uh if it's unpredictable then optimizing should take that into account but it does look like people are deviating from what they would do optimally trying to take the long term into account uh and taking into account the actual amount of uncertainty um that seems so for example there's something called discount rates in economics uh and interest rates and so uh if you have the long view you would take a very low discount rate that is you you would think the future was just as important although with uncertainty that might make it hard to make some other choices but for example resources are just generally useful so if you just invested in collecting resources you don't have to know exactly what's going to happen in the future to know that resources are going to be pretty useful whatever happens and so if you are discounting the future and not being interested in collecting resources that's a pretty clear indication that you're not so interested in the long-term future relative to other people who might but if it's true that like uh thinking about the long-term future is useful then wouldn't you expect the most powerful institutions today to be the ones that thought about the long-term future in the past well that's uh part of the question i raised in my discussions of it so you might think that in the very long run there would be selection for thinking more about the long-term future because that would win if we look uh at biology we can think that there's been a very strong competition in biology for influencing the future and in some sense that is what evolution is about species and individuals compete to influence the future like having more descendants is the main way that biology influences the future now uh animals sometimes make plans and uh you as a human might make a plan to save up resources and pass them on to your kids and we know that uh that sort of planning has a discount rate that goes with it because your kids only share half your genes so your g individual genes they can't save resources for themselves they can only save resources for this whole creature that you are which is a bundle of genes and this whole creature that you are when it passes on and has descendants those descendants only have half as many genes and so that means that uh your genes have a you know choice between you know doing things that promote the genes right now versus doing things that collect resources and have a 50 50 chance of promoting that gene later on because its descendants may only have a 50 chance of sharing that gene so that produces a discount factor of once of generation a factor of two and that's in the ballpark of what actual human discount rates are and so that is a plausible explanation for human discount rates is that uh we evolved in this context where uh when we had a choice to save resources in order to pass it on to the next generation uh the next generation only shares half our genes so see the challenge is to come up with a unit of selection that can pass on resources to itself over the long run and so what that might be perhaps is a nation or a firm or religion or some organizational unit and i think we should reasonably expect that in the long run organizations will find a way to collect resources and then promote themselves in the long run but they haven't done that so far because mostly organizations don't really have so much of a separate existence in our world they are a way that our genes or our individuals use genes use individuals and individuals use organizations to further their ends but if the organization wants to just take over and says i don't care about you jeans or individuals i just want to promote myself then the individuals and the genes will rebel and you know bring the organization back in line and that's kind of what's happened so far organizations haven't really been able to promote themselves as much as they might like so the closest you might get is say the religions encouraging proselytizing where they're promoting the religion right but they found a way of doing that via the individuals proselytizing promoting themselves yeah and so they found a win-win deal where the organization can promote itself and the individuals can promote themselves but you know that's hard to find and so organizations haven't been able to just collect resources because what happens is if a church collects resources or a firm or club well later on some humans are in charge of that organization then they just take the resources they just grab them they say the heck with this organization those are tasty resources i can promote myself and my genes by just grabbing their resources and taking them away from the organization so you know for the long run the question is how can organizations find a way to collect resources and keep it directed for the purpose of the organization as opposed to allowing the people who run the organization to grab those resources for themselves have you considered doing this yourself raising some money and then like building an institution to shepherd and distribute it in the future well i i mean i've fought it through just in terms of the just understand what's going on i'm not you know i'm not rich i don't have that many resources but one of the big things that's happened is this might happen if we used contracts to make it happen so for example i might make an organization and then i make the constitution of the organization a certain way and then i say everybody who wants to join this organization you have to agree to this constitution and then we're going to use law to insist the constitution be obeyed and if we if we were allowed to write a a constitution that specified saving resources and that constitution would be enforced and then the organization was just not going to be taxed or destroyed then this would work and what's gone wrong in history is that when organizations tried this people associated went to the courts and said no no we want those resources and they push the courts to interpret law so that these con these contracts are not enforced so we do not allow arbitrary organization contracts that allow the organization to save resources into the future that's specifically something that law has prevented what you just described sounds like a nation state so is it a nation state the unit that's figured out how to act as like an evolutionary agent and grow more powerful no because nation states can't just collect resources and save them the people who run the nation state they get to spend them so in a democratic nation state for example if the country collects a bunch of resources then the voters go tasty and they can vote in a politician who will vote to spend that you know will choose to spend those resources on the voters and not save it for the future similarly even even when there's a dictator the dictator doesn't set it up forever for the organization they're setting it up for themselves or perhaps their children and again they don't set it up to just keep collecting resources but so so just let's walk it through today and for a very long time almost you know as long as we know interest rates have been higher than gross rates that as rates of return on assets have been higher faster than the economy grows so if the economy you know doubles in 20 years and assets double in 10 years you see then any organization that chooses to grow assets will grow as a fraction of the economy until it comes to dominate the whole economy and then at that point interest rates would fall until this organization really couldn't get a return above the growth rate of the economy but that's an opportunity that's been sitting there for a long time and not really successfully taken advantage of nobody's really managed to create organizations that just keep lasting and collecting resources growing faster than the economy and therefore growing as a fraction so what happens well either the people running the organization decide to take it or outsiders decide to take it you know the local city or the local government says tasty and they decide to tax it or or just grab it or in war uh invaders come and take it and so consistently over the long run um we just haven't been able to do that but it's a thing that remains possible and i predict that eventually there will be some sort of unit of selection that will take the long view that will be able to collect resources and make choices and promote itself uh with a long view in mind and that when enough of those units exist then they will drive interest rates down to not exceed growth rates and then at that point the world will be full of creatures who take the long view and then long term will not be neglected so today you may well fear and and worry that there are things that will happen in the long run that could kill us all that will destroy us but nobody really takes the long view so nobody's very interested now and working to prevent and uh that's because we just don't have any units who take the long view individual humans don't because we have these discount rates organizations don't because they are run by the humans uh you know and other biological species they really can't even plan and think about the long run so you know that's it but as you describe if religion implicitly at least acts as if it's uh planning for the long term in some ways but not others so that's the key thing to do organizations are not like agents who can plan everything with autonomy they're just a collection of strategies that humans use and in some ways they can manage to promote themselves in other ways not but that they aren't these conscious agents planners who you know have a utility function and can make complicated plans to achieve the organizations that but isn't that exactly what you'd expect in like sort of an evolutionary landscape just like this sort of eventually right so that's the whole point eventually but you know organizations are new i mean the key point is you know humans have been around for a long time humans with organizations you know initially for a million years human organizations were like 50 people and that was uh quite a distinct thing from other primates we took a lot of work to agree able to make human organizations with 50 people and have them function and those were human bands and then with farming 10 000 years ago we managed to make like organizations on the scale of a thousand people for a village and then maybe empires but empires only control a very limited range of things that people do in villages so for the most part empires were just things that handled the military and if the if an empire came by and they wanted your sons for soldiers and they demanded taxes well you gave it to them but otherwise they didn't really run things sorry so i was i meant like isn't that exactly what you'd expect in the long term isn't like you'd expect these unconscious basically memetic um units to succeed in the long term meaning even in the future we might just have these like cultural units succeeding and not in org what's their say that organizations are more successful you need a unit that can plan in order to predict that a unit will successfully plan to promote itself right so if you just talk about a meme well a meme is just one tiny piece and it really can't promote itself except in conjunction with a bunch of others that form organizations just like genes right genes really don't promote themselves by themselves genes promote themselves via organisms and organisms are big complicated collections of genes because individual genes they can't plan they can't store resources they they can't make strategies you know they're very small pieces it's the whole collection of genes that is capable of doing those things and similarly for organizations you know they have to be have to have a lot of parts that all work together to make an organization work and uh the question is can you know can you do that so tyler cotton writes and stubborn attachments that if you really care about the long term the best you can do is just promote economic growth uh do you agree with that uh well i mean it depends on what you care about the long run so if you only just care that everybody's happy in the long run regardless of whether you're allied with them or whether you like them or anything like that then you just want the future to be big right and economic growth is the future being big right so like i mean that's almost sort of by definition right if you say well i don't really care who's in the future or what they're doing or or what's happening i just care that there be a big future and there will be a lot of things happening well then you won't grow because growth literally is the measure of like how much stuff is happening right how big it is right so the question is what do you care about now when i predict there will be organizations with a long view i'm not predicting that kind of preference because that's doesn't you know that's not very stable to to competition so when we think of say biological organisms having a long view well they have a long view to promote themselves they have a long view to to create descendants that's not the same as wanting the entire earth to have more biomass or something that's them wanting to continue so that's to me though the easiest thing to predict is that eventually there'll be units of selection organisms uh organizations that take along you view with respect to themselves yeah but the rivalries that exist today you wouldn't expect them to exist in the far future so why would you even consider them when you're uh planning for the future why won't you just try to lift all votes well that isn't selected for it you can if you want i mean you you know you don't have to choose to have yourself represented in the future you can just choose to promote the future if if you can and will then you can just do that right but what we predict is that the future will end up being dominated by creatures who do promote themselves and who take a long view about how they promote themselves that's the thing we should be able to predict the future has so uh you know you have a choice about whether you want yourself reflected in the future whether you want to be part of a unit that's competing like that and taking the long view but if i mean so say you take a long view but you don't promote yourself well whatever it is in yourself that makes you take the long view that's dying out you aren't taking the long view on the long view in that case you're having a current long view that you're acting on but you're not acting to preserve the long view that you are yeah and have it last if you want a long view that lasts then you need to focus on preserving a long view on promoting the thing that embodies the long view and having it last yeah that makes sense okay let's talk about the elephant in the brain um so you you split uh you explained there that we have hidden motives where we actually want one thing but we pretend we want something else um our subconscious and conscious intelligence says different like our capacity to actualize our implicit goals and then our capacity to justify them explicitly so let's back up so we make sure we don't confuse people um we are complicated creatures and there are many sort of levels at which we make choices and so we can talk about causes of our choices that are approximate that are close to the choices and distal far away from our choices and we can have different ways we think about the causes distal and proximate so that's the key thing here so when we are talking in our book the elephant the brain about motives we're focused more on distal motives we're focused more on from a distance what are the sort of fundamental forces that are pushing choices one way or the other those forces may or may not produce different sort of conscious thoughts just before the choice is made so if we think about proximate choices you might ask well just before i made the choice what was going on in my head what ideas did i have what was i talking about what would i have said if you had asked me about it those would be proximate causes those would be the things closest to the choice but those are complicated and context dependent and in many ways i think it's more robust and even easier to focus on distal causes what the fundamental forces are that are driving choices one way or the other and those forces can reflect themselves differently in different final choices or context so for example in different times and places in history people have thought differently about different things and they said different words about it and they you know different thoughts come to mind but often it's the same fundamental forces back behind all of that that push people to do things you could take romance for example you could say you know the idea of romance and what what appropriate things are for relationships and you know which kind of relationships we approve of and when you're thinking of entering relationships what's in your head and what are you saying about it now those things have changed a lot through history in places right people have thought very differently and they're changing them now but if we stand back and say why do relationships exist well it's a simpler theory it's a simpler story well we get why relationships exist because you know that's how biology reproduces without relationships there would be no descendants and so from that level it's relatively easy to find this robust explanation uh for you know some of the major elements at least are relationships that are independent of time and space and are relatively simple so similarly in our book in the brain we're talking about sort of the more basic fundamental motives behind many things we do the mode of being just a mo a force a and a you know a thing that produces it and so we're less interested in what's in your conscious mind at the time and whether you're aware of it and what you would say about it and maybe even how much that varies across nations across time across social classes so because those are all interesting but i think the first priority would be to just sort of get the common elements like if there's a common element behind schools or medicine across time and space and social class we should probably know that first before we try to figure out how you know rich people do it differently from poor people or europeans do it different from asians or it was done differently now versus a century ago i mean those are all interesting to some extent but you know the first cut should just be what's the common element but so if we go along that demarcation the question then is is our capacity to realize our basic motives is that correlated with our ability to explain them later on with our conscious mind well so the key idea in our book is to say your conscious mind has a purpose it's built for a reason but it's not the being the king or president that you think of or perhaps have been told that's not what your conscious mind is for so your conscious mind is more a press secretary its job is to watch what's going on and come up with good explanations that make you look good and in particular that protect you against accusations of norm violations and so your job isn't to know why you're actually doing things your job good reason to explain why you're doing things but then so my question is um is there some sort of correlation between how good the president is and how good the press secretary is because people have differing abilities to justify their actions and people also have differing abilities to actualize their base motives is there some sort of correlation there i mean there probably is just because there just seemed to be this overall correlation and abilities of all sorts you know iq and wealth and you know just good things tend to be correlated so i would expect they're correlated roughly not not finally but roughly just because you know people who are more capable a tend to be more capable of boo b for almost all a and b that's just a general thing for people okay uh would you expect meditators to signal less to just be more upfront about what their motives are oh no could you tell me why you would expect that um they're more self-conscious so that they can't dilute this they maybe they can't fool themselves as easily and so they're less likely to fool others well first of all i'll ask how do i know they're more self-conscious or self-aware as i think you mean self-aware yeah they're more aware of what they're doing and why um i don't know that they are so so let's talk about sort of the truth oriented community what does that mean so a lot of organizations and groups and you know people in the world tell you that they are part of the truth-oriented part of the world and that's part of their identity they tell you well those other people there are deluded but we we are into truth it's one of the favorite worlds and say among christians use christians are telling you we care about the truth we don't you know we have a strange religion but it's true and we are into truth and those other people are not into truth so just everybody likes to think of themselves as the people who are into truth and who are less diluted than the other people so i just can't take people at their word for saying that they are less deluded they might be but merely the fact that they like to say that about themselves isn't enough for me but i want to see some concrete evidence that they are actually more truth-oriented but you have no reason to suspect that a practice like meditation would reduce the distance between your uh you know who you present yourself to be and who you actually are or well then i'd have to dig into what i know about meditation now you may well know a lot more than me i don't yeah i i will i could defer but um the last time i did a bunch of reading about meditation was just before my interview with sam harris because he was in meditation we never ended up talking about it but but reading about it uh to me i saw this focus on sort of um a certain sort of meditative state of mind so it's they seem obsessed really with a certain sort of state of mind they are really really trying to achieve that and they see the that's good for themselves and they somehow think it's good for the world and right there that looks a little deluded i gotta say you know because whatever the state of mind they seem to get into i'm not seeing much evidence that it's like this thing that helps the world a lot i'm not even seeing that much evidence that there's a thing that helps them a lot what i see is that it's very prestigious in their world they you know that is their ranking pecking order the pecking order is who has achieved this higher level of consciousness in their meditation and okay i mean every world has some sort of pecking order and why not that one but uh i don't see much of a practical use of it really compared to all the resources they're investing in it yeah i mean they put a lot of time and trouble into this thing which have a lot of opportunity like people take a decade off to go meditate or something like just so later on they're going to be better at something i mean that there's almost nothing you can do for a decade which will somehow change you in a way to make you more productive later on i think a decade is like huge yeah uh you know the opportunities cost is enormous so so right there some of the specific things i know about meditation they don't give me much confidence that these people are less polluted on average than other people because you know they're into us a relatively religious kind of frame of mind where they've got a particular pecking order they tell themselves that no no i'm not doing this for me i'm doing this for the world and consciousness and whatever and i've i've met people who you know who get together and think well the way we're going to create world peace and the way we're going to make innovation the way we're going to make world better is they're gonna go meditate for a long time you go what should feed the poor or like stop wars or no you're gonna go meditate like okay so let me they might even be right i mean i could be wrong about this but you can see like where i'm coming from in terms of on the face of it there's a reason for skepticism yeah no i don't meditate uh but let me let me try to steal man their position so the claim is you're more aware of the next thought that enters your mind and because that awareness you can choose you can realize that your own motivations are uh false to some extent and so you're less likely to project that false motivation to others do you buy that chain well so we wrote this book the elephant in the brain that takes ten areas of life and talks about how the letters are different if you showed me a book on meditation that went area by area life and showed me specifically how they had figured out that the motives were different i would be impressed if they just say i can see that my motives i'm lying but and i say well what exactly are the lies you know so lie is where you say a and b is really true if you could make a long list of the specifics a's and b's i think you know i i believed a but b was really true and that applied to a lot of people well that would be impressive right you show me the meditation handbook that gives me those long lists of a's and b's and i would then say yep those people seem to get uh you know self-deception and hidden motives without that list of a's and b's you know lots of people in the world like to say i see truth and they like to say that uh once upon a time i was deluded and then i saw the truth yeah and they don't like to get in the specifics of what exactly they are deluded about other than the fact that they weren't they had yet to join christianity or yet to join effective altruism or whatever it was they were the usual thing they're happy to tell you about being deluded about is not having joined the new group they're part of now if i want a longer list of other things they were diluted by that's sort of missing so robin hansen won't be meditating anytime soon um i might but i mean i'm not going to put this huge glory priority on it like i i scratch my back sometimes it's nice a meditation can be nice too but it's like i don't say scratching my back i'm doing for the world or it's going to create world peace or we're all going to have a unified world together or scratching our back no scratching your back is just nice that sounds like a strawberry but robin wright actually wrote a book where he claimed exactly this um your back makes world peace okay so along with the faculty to like fabricate pro social reasons for what we're doing uh why don't we have the capacity to uh consciously detect other people's false uh explanations of what they're doing that seems kind of useful well we do um when we have rivals we are unusually able to sense when they might be hypocritical and when they might be lying about their motives of course we go even farther than that and attribute falsities and delusions to them that they don't have because we're really eager to trash our rivals what we have a harder time doing with of course is noticing our own things and things in common so uh in our book we talk about you know many kinds of hidden motives and some of the hardest ones for us to see are the ones that we all share so for example you might say that your rival doesn't care about their grandma because they won't take their grandma to the hospital and the rival doesn't care about their country because they won't pay for more national health care for the rest of their country members but if all of us are deluded about medicine actually being helpful that's not going to show up in those sorts of accusations about rivals because those are sort of appealing to shared senses of what's important and then saying this particular person doesn't share what the rest of us share about what's important and so um you know the more that motives vary across people the more you're going to be able to be able to notice that a rival might be having a different motive than they say but when motives are more common uh that's harder to see why have we even retained social norms why does society even evolve to have the dorms it has if uh if it's so easy to skirt them by just like pretending to uphold them it's not that easy so i'm just just be clear for our listeners um humans are unique compared to other animals and having norms that is rules of what people should or shouldn't do that we enforce via collective groups that is if i see you breaking your rule i'm not just supposed to do something about it i'm supposed to appeal to other people around me to tell them that you violated the norm and discuss with them what we should do about it so that we collectively punish and if i see you violate the norm and i fail do that and somebody else notices that that's a violation on my part so that's the key idea of norms and norms are what we have before law to uh make people behave and norms are still very important and in some sense norms are more flexible and fluid and so we use norms when they are seen sufficient and we resort to law for things that norms don't seem up to the task of in our modern world of course up until let's say 10 000 years ago there was no law norms norms were everything that's how we dealt with all problems between people was norms um so norms are still this you know superpower humans have and we continue to use norms we've augmented it with law and the context in which we enforce norms has changed in the sense that we have much larger communities of talking so you know in the old world of a small band i saw you break a norm i could tell my friends they could tell their friends and pretty soon like all 30 of us have hurt and we're done we've all gossiped about it of course if you have counter argument you've gossiped your camera argument and maybe i'll get around the campfire tonight and talk it out and you know make a decision and in our larger world we have troubles with that because you know we're in larger communities and we just can't spread the word about everything everybody and uh we have to decide who to tell what and people have to decide what to listen to and what to believe uh in context where they don't know his people as well and that don't have as much context about the accusation and that makes it harder to handle norms in our world but they're still pretty important shouldn't the power shouldn't we be signaling less over time if it seems it seems that like um social norms are getting less rigid over time so shouldn't the amount of signaling go down and the distance between our fabricated and real cells go down over time so i'd say the opposite in the early transition so if you think about foragers say they live in a group of 30 people these 30 people have been with each other their entire life they know each other pretty well so whenever they hear about something somebody's done or an accusation what they've done they can put that into a lot of context now those 30 people have been spending their entire lives trying to impress each other but in any one moment they don't have to make sort of a crude signal that you that it's hard ambiguous to interpret you know this person pretty well and now you are slightly adjusting what you know about them in the context of this thing they did yesterday but um that's much more sort of integrating the information about what they did into a solid impression of course people don't change that much over a long time so mostly you just keep reconfirming what you already knew about their abilities and inclinations right once you moved say to a village-sized world for the farming world of a thousand people now you didn't know everybody that well and now any one thing somebody did is much more of a potent signal about them and so i think people paid a lot more attention to crude and potentially misleading signals once uh they're a lot more ignorant about the person who the signal is about so in a you know a group of a thousand people uh you know somebody comes to you and wants to marry your daughter well you haven't been around that person your entire life you you can use your connections to talk to other people who maybe have and try to get information that way or if there's an accusation that somebody's done something bad you have to decide whether to support that but you don't know that much about this person you hear what they said and who's accusing them and so in some sense signaling matters more because you know less but you still have to make key important decisions and in our modern world you know you could see that continuing on i mean there are many people we suddenly read about in the paper we've never heard about before ever in our life and we're making judgments about them and so we have to make these crew judgments based on where they live or what degree they have or their gender or their age and so therefore in some sense signaling becomes very potent in that context all of a sudden say the world is accusing somebody of something you have to decide whether you're against them or for them and all you have is a small number of clues and so they've known their whole life that they need to try to make those clues look good so that at this moment if it should ever come that you will be favorably inclined toward them but isn't there another sense in which each person matters less and more people you have if there's only 30 people in a tribe some young boy is very likely to marry your daughter but if there's thousands of people the next guy you meet is very unlikely to marry your daughter so you're less likely to be observing the signals or displaying signals to the random person because they're less likely to be important to your life but a lot of the signals you say it doesn't cost anymore to send it to the to any one extra person that you've sent it everywhere you know so you you build up your muscles and then they show up and anybody who happens to see you can see your muscles right and you have a beautiful singing voice and anybody hears your voice can hear it and you know you have wealth and you've got gold jingling on your chain on your wrists and anybody who sees you can see that and so in a lot of ways your signaling isn't targeted at particular audiences uh you are just collecting signals that are seen by a lot of people but what about signals that are powerful only because they're targeted like narrow casting well so that's a very different sort of thing so loyalty signals i would think are much more of that and so when people think of signaling they mainly think about signaling ability or capacity but in fact loyalty signaling matters about as much as ability signaling and loyalty signaling is about who you're loyal to in what ways and of course that's a signal about a particular or other people right so now for example i say that medicine is largely a loyalty signal that is you spend money to help somebody else get medical care and that's a signal to them that you care about them of course you'll have to target that at them uh and but you know the question is how much more do they know about you uh in order to interpret the signal so the better they know you the less they're going to miss to badly judge you on the basis of this one signal but even the even when people know a lot about people they still put a lot of weight on things like this so i mean honestly if if you are suddenly injured and you are feeling really scared and you really are scared of your life and whether you're going to be taken care of having somebody known all your life likes to quit work for a bit coming to take care of you like be around you all the time say nice say things that feel that means a lot to you so clearly in some sense even if you've known this person all your life you aren't making further inferences about them in that context and there's a sense in which if you've never felt at risk of your life and this is the one moment when that happens you've never had that signal from them before and this is your chance to get it are we signaling more now than ever because of like social media that's not obvious we're signaling to different people in different ways i'm not like you know honestly i would say you know 80 of what we do with signaling in some sense and that's been true for a long time and social media just moves it around huh but if we're if we're consciously like putting up an image on instagram or snapchat isn't that that must increase the like the effort we're putting into signaling your ancestors were doing a lot of things consciously too i mean your ancestors you know bought a house and they wore certain clothes and they went to certain meetings and they made sure the fence was painted and you know they were polite when they talk to people i mean almost all people always have always been doing most of the things they do with an eye toward how it will look to other people okay interesting uh i wanted to ask you about your um chapter on conversation so you explained that conversation is actually i'll let you explain first what what conversation it's for well so i mean we haven't said that much about the book so let me just say oh yes the setup of the book is the first third is making it plausible that people might not be aware of their motives and then the last two thirds of the book goes over ten different areas of life and for each one it says in this area of life if you ask people what their motive is this is what they will say here are a bunch of puzzles that don't fit that very well and here's our better explanation of the motive that's really going on which is hidden one that people don't usually admit to and conversation is one of our chapters and so in that structure the first thing we have to ask is well why are people talking now we want to set aside very practical talks so if you're ordering a pizza or you know testifying in court there's a lot of context in which the purpose is pretty straightforward and mechanical in the context it's in we're asking about conversation so as you know we're often just talking without much of a purpose in mind so what we're asking is what are you doing there when you're shooting the breeze chatting uh talking about a wide range of topics that aren't very directly related to anything you're doing so if we ask you what are you doing when you're talking because you're doing a lot of talking the most common explanation i think you'll get is well we're sharing information i know a lot of things you don't you know a lot of things i don't if we just share back and forth things we know then both of us will know more and that will be great and that's not crazy of course all of the things we say as motives aren't crazy they do apply some of the times that's why they work as an excuse when we say they aren't our real voters what we mean is that's not our main most common motive it applies less often than we admit so yes we do exchange information and that is useful but it's not the main reason we're talking and how can i say that well there are these puzzles that don't fit so well so if we were exchanging information first of all we'd be keeping track of some debts i'd say i've told you three things useful it's your turn to tell me something useful we'd be less eager to talk and more eager to listen and we'll be talking about important things what's the point of talking about random trivial things uh when you're trying to exchange information but in fact of course what we do seem to we've seen to follow this rule where the conversation is supposed to bounce around at random and wherever it goes you're supposed to have something to say you're not supposed to try to make the conversation go where you want it especially not to important things and we do try to talk more than we listen and we don't keep track of debts so what's going on what is the better explanation of conversation so our story is you could think of yourself as having this mental backpack of tools and resources every time a new topic comes up your job is to find something in that backpack that's relevant and interesting and even impressive but more consistently and often you can do that play this game whatever subject comes up you've got something relevant and interesting the more impressive that backpack is that means if i'm your friend an associate and some problem happens to me and i tell you about it you're going to use that magic backpack to bring out something useful and that'll pretty be helping me and so the ordinary conversation is the game where we test each other's backpacks so we check to see well how useful is it how full is it what do you got in there so that we could get we could say that we want to get access to it later um just because they'd always be around okay but but the fact that we want to talk more than listen uh isn't that a problem for this theory that we're just looking for potential mates and allies because we'd also want to listen to the signal our potential mates and ours are sending us so shouldn't we want them to talk about as much as we do anyways well we both we always both want to listen to other people's signals and send our own uh but of course we are usually more interested in ourselves than other people and so uh you know often we think we've roughly got a pretty good gauge of them and slight variations on them aren't so important to us but slight variations on us are much more important we want to edge ourselves up just a little bit more uh so that we can seem to be better you know if you thought of somebody who spent you know who was a really good judge of other people but didn't look very impressive themselves versus somebody who looks really impressed themselves but who doesn't seem that good a judge of other people who are people most impressed by yeah that makes sense but just as a just as um a tool in judging people's uh i guess just in terms of how useful it is shouldn't we expect people to also care as much about what what other people are like because if you're talking one-on-one it matters just as much if like you care but the question is just on the margin how much extra effort are you going to put into it right so i mean it's often a fact when people are talking they can either listen to what the other person says that they can be planning the next thing they say and it's just a matter of the relative priorities there if you think you've got a pretty good judge of them then you may just want to focus on moving yourself up another way you might think about it is for each person there's a threshold of whether they're good enough for you and what's mainly important is whether they're above or below that threshold uh and it might be relatively rare that they're right near the border of the thrush thread where you're really trying to make a judgment about whether they're good enough you know sometime on first date say or interview for hiring now now you are trying to they might happen to be near the threshold and then it's really important for yourself though there's no threshold effect you just always want to be better right because you never expect any one person you might be good enough but there's you know they might talk about you to other people and you just you just want to go up okay i see that makes sense uh oh so you wrote in that chapter because you're saying that conversation is mostly a way to judge people's capabilities you you write in that chapter um we do not routinely expect mumbling uh stuttering scatterbrained politicians so uh you're gonna have to explain the nominees of the 2020 election to me uh well i mean there's no doubt that relative to a median citizen the two presidential candidates are impressive they you might think there's even better options out there but now you have to look at all the other constraints so um you know leaders have to meet a lot of different constraints uh that that you want to select for leaders and politicians like democratic leaders have even more constraints and then there's a lot of randomness going on so again you know our book and our analysis is focused much more on the average case than on individual deviations uh but you know both both presidential candidates i understand at a rough level how they got there and why they're there and often it's a matter of compromise between different groups uh if you can separate groups into uh distinct groups each group has a sort of their favorite person that's not necessarily who goes up it's often a compromise between different groups and that compromise isn't necessarily the best from any one group's point of view you know if you just want to be an impressive speaker you know having a simple absolutist position that you take to the extremes is much simpler and cleaner and clearer and uh to explain but of course that's not where the median voter will go political processes when to choose compromises and compromise positions aren't as easy to explain or rationalize or inspire people with and so there's there's a trade-off there between getting a person who sits in the middle of a large group who everybody can you know at least not hate too much and having a person with a simple clear message that resonates and inspires people with the message oh so are you saying that being articulate could in some ways harm you as a politician because oh yes so i mean uh you know there's a standard observation about politicians is that they are not very specific with their positions they are often purposely ambiguous and they often seem to speak different things to different audiences the ambiguity is part a way to cover for that is to say no no i didn't really say different things because these words cover different purposes but yes politicians uh if they were very clear about their positions that would be a big turn off to a lot of people for whom the clarity is not the position they want so if they can be vague the vague words can encompass a lot of possibilities uh and you know that's only articulate in a certain sort of politician-y sort of way if you look at politicians speeches you know there's a certain style in which they're articulate but they are it's a very ambiguous style it's a very not clear and precise style in which they are inspiring they're supposed to sort of throw out words that are emotionally lit you know loaded and uh and evoke images in your mind without committing them to that much specifics but uh if you think of somebody like bill clinton he was somebody who was articulate but was also you know many people thought that he represented them with the current nominees it seems that they don't have the same faculty of clarity and conversation um no doubt it varies although you know the word articulate is often a class loaded word so for example you know uh trump went out of his way to not project high class you know associ signals his winning uh in 2016 was basically saying there was a group of people out there who felt that they were being ignored and he basically credibly signaled i'm one of you and i'm not with them and so therefore i will represent you in order to do that uh he had to show an affiliation with them different than the others and so he had to be a contrarian signal in some ways he if he just gave the usual sort of political prestige signals that would not convince them that he was one of them and since they felt that they were on the out so one of the things they felt is that they were being ignored by cultural elites and so trump needed to not be a cultural elite uh in order to say i'm with you and not with them and you know they were also more rural and uh you know working class and things like that and so trump explicitly purposely successfully said i'm one of you by choosing cultural styles that were like them and specifically different from the others and in ways to make the other people yell at them you know what better way to convince you that i'm not one of them is to make them yell at me then then i'm not with them i'm with you this all makes sense but it seems to contradict your uh explanation and the chapter on conversation if if conversation is a way to judge whether we want somebody in leadership either just not important at all or or so conversation everything we do i mean importantly no everything we do has many motives that are relevant so uh there's the most the one we'll most often mention and then there's the one that most often matters but there's always a lot more so in many kinds of conversations loyalty signaling is just as important as ability signaling right that's not so much in one-on-one conversations although it's important there it's important for example when your friend complains about something that you show support that's a loyalty signal you're not being very impressive with your support you're just being loyal with your support that you say yes you know when you did the interaction with them they were wrong and you were right and we do those sort of loyalty signals all the time when people around us complain or indicate some sort of rivalry and that's a very common thing we do and of course in politic
Original Description
Robin Hanson is a professor of economics at George Mason University. He is the author of The Elephant in the Brain and The Age of Em.
Robin's book: https://www.amazon.com/Elephant-Brain-Hidden-Motives-Everyday/dp/0190495995
Robin's website: http://mason.gmu.edu/~rhanson/home.html
Robin's blog: https://www.overcomingbias.com/
Episode Website: https://www.dwarkeshpatel.com/p/robin-hanson
Apple Podcasts: https://apple.co/3Rmz9MB
Spotify: https://spoti.fi/3Rm7RWS
Follow Robin on Twitter: https://twitter.com/robinhanson
Follow me on Twitter: https://twitter.com/dwarkesh_sp
Timestamps:
00:00 Intro
00:05 The long view
15:07 Subconscious vs conscious intelligence
20:28 Meditators
26:50 Signalling, norms, and motives
36:50 Conversation
42:54 2020 election nominees
49:25 Nerds in startups and social science
54:50 Academia and Robin
58:20 Dominance explains paternalism
1:09:32 Remote work
1:21:26 Advice for 20 yr old
1:28:05 Idea futures
1:32:13 Reforming institutions
Watch on YouTube ↗
(saves to browser)
Sign in to unlock AI tutor explanation · ⚡30
Playlist
Uploads from Dwarkesh Patel · Dwarkesh Patel · 4 of 60
1
2
3
▶
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Rubik's Cube Encryption Demo
Dwarkesh Patel
Bryan Caplan - Nurturing Orphaned Ideas, Education, and UBI
Dwarkesh Patel
Matjaž Leonardis - Science, Identity and Probability
Dwarkesh Patel
Robin Hanson - The Long View and The Elephant in the Brain
Dwarkesh Patel
Caleb Watney - America's Innovation Engine
Dwarkesh Patel
Alex Tabarrok - Prizes, Prices, and Public Goods
Dwarkesh Patel
Scott Young - Ultralearning, The MIT Challenge
Dwarkesh Patel
Scott Aaronson - Quantum Computing, Complexity, and Creativity
Dwarkesh Patel
Uncle Bob - The Long Reach of Code, Automating Programming, and Developing Coding Talent
Dwarkesh Patel
Michael Huemer - Anarchy, Capitalism, and Progress
Dwarkesh Patel
Sarah Fitz-Claridge - Taking Children Seriously | The Lunar Society #15
Dwarkesh Patel
Byrne Hobart - Optionality, Stagnation, and Secret Societies
Dwarkesh Patel
David Deutsch - AI, America, Fun, & Bayes
Dwarkesh Patel
Bryan Caplan - Labor Econ, Poverty, & Mental Illness
Dwarkesh Patel
Jimmy Soni - Peter Thiel, Elon Musk, and the Paypal Mafia
Dwarkesh Patel
Razib Khan - Genomics, Intelligence, and The Church of Science
Dwarkesh Patel
Pradyu Prasad - Imperial Japan, the God Emperor, and Militarization in the Modern World
Dwarkesh Patel
Manifold Markets Founder - Predictions Markets & Revolutionizing Governance
Dwarkesh Patel
Ananyo Bhattacharya - John von Neumann, Jewish Genius, and Nuclear War
Dwarkesh Patel
Agustin Lebron - Trading, Crypto, and Adverse Selection
Dwarkesh Patel
Sam Bankman-Fried - Crypto, FTX, Altruism, & Leadership
Dwarkesh Patel
Alexander Mikaberidze - Napoleon, War, Progress, and Global Order
Dwarkesh Patel
Sam Bankman-Fried On FOCUS
Dwarkesh Patel
Sam Bankman-Fried on GREAT FOUNDERS
Dwarkesh Patel
$30 BILLION Opportunity Ignored by Sam Bankman-Fried Competitors
Dwarkesh Patel
Fin Moorhouse - Longtermism, Space, & Entrepreneurship
Dwarkesh Patel
Joseph Carlsmith - Utopia, AI, & Infinite Ethics
Dwarkesh Patel
Will MacAskill - Longtermism, Effective Altruism, History, & Technology
Dwarkesh Patel
Steve Hsu - Intelligence, Embryo Selection, & The Future of Humanity
Dwarkesh Patel
Austin Vernon - Energy Superabundance, Starship Missiles, & Finding Alpha
Dwarkesh Patel
Charles C. Mann - Americas Before Columbus & Scientific Wizardry
Dwarkesh Patel
Tyler Cowen - Why Society Will Collapse & Why Sex is Pessimistic
Dwarkesh Patel
Bryan Caplan - Feminists, Billionaires, and Demagogues
Dwarkesh Patel
Brian Potter - Future of Construction, Ugly Modernism, & Environmental Review
Dwarkesh Patel
Kenneth T. Jackson - Robert Moses, Hero of New York?
Dwarkesh Patel
Edward Glaeser - Cities, Terrorism, Housing, & Remote Work
Dwarkesh Patel
Byrne Hobart - FTX, Drugs, Twitter, Taiwan, & Monasticism
Dwarkesh Patel
Nadia Asparouhova — Tech elites, democracy, open source, & philanthropy
Dwarkesh Patel
Bethany McLean — Enron, FTX, 2008, Musk, frauds, & visionaries
Dwarkesh Patel
Holden Karnofsky — History's most important century
Dwarkesh Patel
$30m Grant to OpenAI?
Dwarkesh Patel
Does GPT Have Holden Worried?
Dwarkesh Patel
Lars Doucet — Progress, poverty, Georgism, & why rent is too damn high
Dwarkesh Patel
Deep Learning Changes Everything
Dwarkesh Patel
Garett Jones — Immigration, national IQ, & less democracy
Dwarkesh Patel
Marc Andreessen — AI, crypto, 1000 Elon Musks, regrets, vulnerabilities, & managerial revolution
Dwarkesh Patel
Why You Shouldn't Start A Startup
Dwarkesh Patel
The Future Of Venture Capital
Dwarkesh Patel
The Crucial Skill For A Startup Founder
Dwarkesh Patel
Brett Harrison — FTX US former president speaks out
Dwarkesh Patel
Nat Friedman (Github CEO) — Reading ancient scrolls, open source, & AI
Dwarkesh Patel
Ilya Sutskever (OpenAI Chief Scientist) — Why next-token prediction could surpass human intelligence
Dwarkesh Patel
Impact of Taiwan Invasion on AI
Dwarkesh Patel
Reliability is Bottleneck on AI - OpenAI Founder
Dwarkesh Patel
Next Token Prediction SOLVES AI Says OpenAI Founder
Dwarkesh Patel
Harmful Uses of GPT - OpenAI Founder
Dwarkesh Patel
Why OpenAI Founder Thinks AI Is Near
Dwarkesh Patel
AI will help us achieve enlightenment - OpenAI Founder
Dwarkesh Patel
Eliezer Yudkowsky — Why AI will kill us, aligning LLMs, nature of intelligence, SciFi, & rationality
Dwarkesh Patel
Richard Rhodes — The making of the atomic bomb
Dwarkesh Patel
More on: AI Alignment Basics
View skill →Related AI Lessons
⚡
⚡
⚡
⚡
The Dark Side of AI: What We Lose When We Stop Thinking
Medium · AI
AI Security Isn't a Product. It's an Engineering Discipline.
Dev.to AI
Why Solving Legal AI's Context Problem Is Harder Than You Think
Forbes Innovation
How Can We Truly Protect Information Privacy in the Age of Artificial Intelligence?
Medium · Machine Learning
Chapters (14)
Intro
0:05
The long view
15:07
Subconscious vs conscious intelligence
20:28
Meditators
26:50
Signalling, norms, and motives
36:50
Conversation
42:54
2020 election nominees
49:25
Nerds in startups and social science
54:50
Academia and Robin
58:20
Dominance explains paternalism
1:09:32
Remote work
1:21:26
Advice for 20 yr old
1:28:05
Idea futures
1:32:13
Reforming institutions
🎓
Tutor Explanation
DeepCamp AI