In this episode, we talk about AI safety, machine learning, meaning, the economist Elinor Ostrom, and more with my good friend Cooper Williams. Cooper is a machine learning engineer based in the RTP region of North Carolina.
Links/more reading:
Elinor Ostrom and her book, Governing the Commons (affiliate link).
Pirate law and the pirate’s code.
Lambda School and their web development program.
Transcript:
hey folks my name is will jarvis along with my dad dr david jarvis we record the podcast narratives narratives is a project exploring stories about progress and what ways are we better off now than in the past are there ways that we are worse off what is the ideal future how do we build it join us as we explore these questions with some of the brightest minds in the world
hey folks today we’re sitting in chapel hill north carolina it’s kind of a chilly day out it’s uh november what’s today’s date i don’t even know late november heading towards thanksgiving i don’t know pretty good day um today i’ve got my good friend coop with me cooper williams and uh coop is a machine learning engineer and you’ve had kind of a interesting career so far i know you’re still young but um you’ve trained in lambda school is that correct is that that’s right yeah so i’m 25. gotcha i think you’re about the same age right yeah 26 right okay cool yeah so um yeah i went to lambda school for data science and i got hired by will straight out of at a lambda school so that was cool because we actually met long before i graduated in fact i think it was a month or two after i moved the area and started lambda school if i remember correctly yeah that’s right we met at a slate star codex meetup that’s right yeah it’s pretty cool um and just for context coop is probably the most talented machine learning engineer i think i’ve met dang man that means a lot oh my god all my other friends are machine learning engineers are not now going to come after me but that’s okay oh boy yeah they got to prove their medal all right well you better start learning jiu jitsu with us exactly that’s great it’s great
so is there anything else i should add on your bio side that people might like to know that might be interesting oh well i don’t know i’m sure it’ll all come out in in the discussion i don’t know if i don’t know i was previously a filmmaker i guess that’s the main thing is uh i was a filmmaker for nine years before doing you know touching anything programming so that’s really what kind of films were we making you know not really films uh so i made some short films in college but the main gist of what i did was a lot of ads a lot of small business ads interesting did not make me very much money um but they they kept me afloat that’s cool um i basically yeah i would basically i was i was like a one man um film crew so i worked for a marketing agency that um that had a lot of small town clients in in oklahoma um although there were some banks in there there was a congressman um yeah and i would just like you know sit down with them and say what kind of video you want and we would just make it and um i sold those videos for pennies on the dollar like i did not know how to sell um because i you know i was always super critical of my work yeah um i i never what i never became the filmmaker i wanted to be and you know i i ended up just making a lot of educational videos yeah um after that because shooting on a set um and setting that up setting that up and like bringing people into to to be my crew that’ll just stress me out so i just started doing video editing after that and that’s actually how i ended up doing computer science was i just got interested in computer science started making educational videos about that and about economics um and then i started teaching myself data science and then did lamb to school and even you’re an instructor as well that’s right yeah currently i’m so on top of my full-time gig with tanjo i am doing a side gig where um i’m a teaching assistant at a company called correlation one and um i’m in their empowerment program which is meant to bring traditionally disadvantaged communities minority uh students my cat is uh attacking the microphone you can toss him off oh it’s okay it’s really pleasant actually okay he is he’s a sweetheart yeah so um yeah so i’m working with ds4 uh so uh correlation one’s a ds4a ds for all empowerment program which brings together uh 500 um minorities uh fellows to learn data science for the first time oh that’s really cool yeah mostly geared towards like a business uh facing a business focused data analyst position um so we’re we’re getting we’re just getting the foundations really strong very cool very cool so i before we jump from here in video production uh and you mentioned you know the product wasn’t exactly what you wanted yes yet have you ever heard the ira glass talk about this he’s like you know the real problem with uh creating art is that um you know the art isn’t great um in the beginning or like as you get better you’re like wow like you know you start producing and then it hurts because it doesn’t feel as good as you know it should be if that makes sense right you’ve got this vision and it’s the reality is just crap next to your vision exactly and like pushing through that is like something difficult but anyway yeah uh interesting so are there this is really kind of practical and i don’t want to spend too much time on this but are there easy gains to be made if like if you’re producing a video is there like where are the the easy wins if that makes sense is it like quality equipment is it um i don’t know planning yeah that’s a good question um so the first thing i would say so the boring answer is audio um you can have a terrible terrible camera and if your audio sounds good then the audience accepts it as being not cheap interesting so like then i would say that’s a prerequisite for making a good video oh wow and uh and i’ll admit i was never that great of an audio engineer never that great of a audio sound really complicated yeah it’s a pain and but the the thing is that if your audio is bad that’s the first thing somebody notices and interesting they never get out of the feeling of watching a video right and never get into it the the fun answer to that is the writing um if you if you just have the creativity and you have a mind to sit down with the you know the person who’s paying for it and say what do you want what feeling do you want to evoke in the video also who’s your audience of course and what is the emotion for them then that’s going to be like where most of your quality is that’s going to be where all the magic happens and your your camera could be crap you know your colors aren’t perfect but at least like you’re if you’re focusing on that emotion you’re going to get something good gotcha that’s super interesting like i’ll tell you an example um so there was a there’s a business coach who i was hired by to make this video um like make a video entry for a competition that was like this international um business coach video competition um and we entered not only in that but in just like a general um like a general i can’t even remember what what it was but bottom line is we entered this competition um going up against um some really big companies and on a budget of three grand i made a video that beat two companies with a combined worth of 7.3 73 billion dollars that way yeah that is awesome so those two companies came in second and third place and on a budget of three grand which is tells you how much i was charging right right right i yeah i beat those guys in an international competition wow that’s it and what do you think what do you credit your success on that was it like taking the time to really you know focus on the problem or what do you think it was well for one thing we we made the video based on a poem that the business coach had written which was just really like i was really unique he put his heart and soul into it and um and it was about business coaching yeah that’s awesome but you know but it had a real charm to it that’s cool we just you know i a lot of the video was um like royalty free footage that i like i just scrounged the internet for a lot of three grand budget three grand exactly yeah that was like my b-roll right artistic b-roll yeah and then like you know um and then i shot a lot of stuff on ou campus which is really pretty yeah and i use like this this cheap um like steadicam gimbal that was only worth 200 bucks and just like really painstakingly like made these steady cam shots um so i basically just you know we just made a plan for what we wanted it to look like and then we did it that is so cool yeah that reminds me it seems like um
people in some sense overrate money you know the money has this like really weird effect on people you know it’s like very people really like it they really want it and it’s but money’s just like you know pure optionality it’s just like you can do whatever with it but it seems like ideas are often times much more important like having good ideas especially nowadays yeah maybe more than ever yeah i agree and skating on that and having some kind of yeah cause a link between your ideas like you know this this um you know this series of images and words is gonna evoke a certain emotion in this audience like you know you’re you’re focused on a certain goal like money is very generalizing it makes you think about uh you know i don’t know you’re just optimizing for getting money which is just like the broadest goal imaginable but if you’re if you want to win a competition with certain audience you know that brings out some creativity built into the to the criteria interesting uh so i wanted to move on a little bit i want to talk about uh an academic that you actually put me on to she’s really interesting eleanor ostrom eleanor ostrom yeah can you tell me a little bit about her and why she’s important and overlooked perhaps yeah well she’s less overlooked nowadays i think um she’s she’s gaining getting uh popularity due to people like tyler cowan got you have made videos about her um and so she is the first woman who ever won the nobel prize um i think recently there was another woman um economist uh her name is she’s french uh yeah uh banner g maybe oh yeah um i can’t remember her first name yeah i can’t remember i think she’s married to a banerjee if i’m not yes yeah and he also won they both won the prize that’s right i don’t know if it’s separate together however right i can’t remember her name god dang it but anyway so oh sorry this is probably pg that’s okay i think i think that passes so so yeah um anyway eleanor ostrom is the first woman to win economics um nobel prize and she struggled against all these institutional barriers to women in economics because when she started it was you know women weren’t allowed literally weren’t allowed in the economics wow and she had to go a different route but anyway her whole scholarship is about the idea of common pool resources so these are resources that are scarce resources usually natural resources but not always um where it’s really difficult to exclude people from using them up um interesting so you can think of this as like um outside of the realm of government uh government uh jurisdiction uh or or should i say uh yeah outside of the realm of um state control and private control it’s like there’s a there’s a third sphere of society when it comes to um when it comes to resource allocation it’s called the commons and so you can so examples of this are like fisheries um woods watersheds you know like an aquifer so these are things where some like someone could deplete the resource if they weren’t careful and so what she studied was why are some of these resources stewarded really well sometimes for over a thousand years without anybody depleting them and some of them are not
she starts she has this great book called governing the commons at the beginning of it um she has this masterful theoretical overview theoretical overview where she basically says um have you heard of the you know the prisoner’s dilemma yeah um so just just to sum it up like um if you have two people who can’t talk to each other and they can sort of they could they could cooperate without talking to each other if they if they did cooperate they would get a big payoff but also the incentives are such that if they cheated against each other then you know they would get a really crappy payoff but they’re both incentivized to cheat against each other right yeah um so she takes that and she says that’s that’s the framework people usually think of common pool resources as being limited by she says this is the huge excuse this is the like number one excuse that um policymakers use for taking people’s stuff and saying you can’t steward this common pool resource we’re doing it for you and so at the outset of the book she says i’m gonna i’m gonna show all these examples where you don’t have to do that and people can govern those resources effectively wow so could you give an example of a a common pool resource that was governed well so i i have an example of one that was governed poorly if you think of like easter island and the trees on easter island so you know they had the the big statues on easter island and um they used up all the trees moving these around and the problem was when you took away a certain number of the trees the wind could start blowing them over and then eventually it’s like you can’t build canoes to get away from easter island and then you know yeah uh so they did a poor job what’s an example of someone doing a good job yeah that’s good governing the commons good question um so there was one uh so there were a number of examples that she gives as well as ones that failed but one example is um a there was a town it was in somewhere in central europe there was a village i believe the swiss alps a village that had a limited amount of grazing area um and yet and and and the grazing area was fragile to some extent and yet they managed to um to allocate the the use of that those grazing resources for thousands for over a thousand years um without sort of depleting the resource interesting um so farmers basically had a system set up to say who can use those fields when and the crucial thing was they had ways of keeping tabs on each other so she said that she says that this is one of the one of the um most important things is that the people who use the resources can set the rules um on how how the resource is used and they can also keep tabs on each other and observe that the rules are being followed so if you can see that the rules are being followed then you yourself you’ll self-enforce gotcha um and so there was no there was no government um like as such it was just it was a it was a horizontal enforcement that’s interesting that almost rides reminds me of the pirate code you’ve heard the pirate code no no so the power code’s really important the pirate code is that if you surrender to the pirates they won’t kill you and you know they’ll just take you back to the nearest port drop you off uh it was really important that all pirates uh maintain the code because if you start killing people when after you took their stuff then they wouldn’t surrender and it’s easier to just have them surrender than actually go and and and fight with people and you know the british navy there you go yeah that that’s that’s smart and that makes sense as a rule um but you do need some like enforcement mechanisms so you don’t have some one person that goes off off the rails and that’s right yeah you you have to have um what she calls gradiated graduated sanctions um and it’s basically a series of um like sort of punishments i guess against against someone who um who breaks the rules and they’re usually enforced on a case-by-case basis by the other people who are using the resource interesting um this is one of seven i believe um guidelines that she gives for or principles actually design principles for common pool resources
so one example of that would be logging in japanese forests there would be these villages where you know you didn’t want to over log like easter island um and so there were set times where people could go out and log and if you were caught by one of your neighbors uh logging in in an unauthorized manner your neighbor would basically demand a certain amount of sake from you interesting that’s funny and and you know and that was how they enforced it on a case-by-case basis oh wow the sanction you know it could be a lot of sake it could be a little right depending on how crappy your family is doing that year economically um but you know um that that basically helped them equilibrate and successfully preserve that resource interesting is this somewhat limited by group size so is there a size where this starts becoming more difficult yeah yeah this is where you get to the dunbar number right okay like 250 people or something is supposedly the number of people that you can know on know personally and have like as part of a tribe i’m not sure how much i buy that and frankly i’m not qualified to to say what the ramifications of that are but it is clear that oh and you know and this brings me to think about some of the other examples she gave in many systems there were thousands of people participating um gotcha so one example would be the um what do they call them ribbon farms in uh in india they had super long farms that were irrigated by um by one stream so they were you can picture like a bunch of ribbons coming off gotcha river and um there were thousands of people and each of those systems was like you know there was a hierarchy of responsibility um each region had its own you know officials those are just councils upon councils sending up their representatives and that all worked really well
uh they um blow it up right yeah um something similar well technically something similar happened with californian watersheds interesting but i i don’t know if you would call that several participants necessarily or like thousands of participants because that was more about organizations um using the water i think gotcha interesting so how did you find ostrom’s work what kind of led you there well i lived with an economist for a couple of years oh interesting so you know and basically when you uh when you live with an economist um you learn to recognize incentives they learn to you know well i don’t know i shouldn’t generalize actually living with this guy this economist yeah living with this particular economist um you know we were just really interested in you know what are the reasons that government tries to take your stuff yeah and think of how we can you know civilly sort of talk the smart people out of that um and you know and and maybe how to you know you know get there get them on our side against the people who don’t know any better right right because there are plenty of those people in power as well definitely so so yeah this is a crucial i think crucially underrated thing in political economy is that the role of the commons the commons is everywhere there are common common pool resources that are intangible such as like neighborhood goodwill things like that interesting um that are just not legible to state actors not legible to policymakers and so they’re overlooked and overrun plenty of examples of this in governing the commons as well very interesting very interesting yeah living with the economist i i recommend it highly recommend it you learn a lot that’s cool so you talked a little bit about uh you know government coming to take your stuff and you know hurt people you know so generally you know our government has a monopoly on violence in this country do you think that’s preferable to other systems is that not preferable what do you think about that yeah that’s a very slight way of asking that question exactly well um yeah i’ve debated whether like how much of this i should i should say on a podcast but you know i basically believe that people have you know individuals have a right to control how their bodies um are used and control how their own property is used and that’s just like the basis of my belief gotcha um i think that there are probably some holes in that argument like some sort of corner cases where you might want to take that away from people but as a general guiding principle i think it works and that does seem to be fairly common sense morality too i feel like most people would tend to agree with that although i could be wrong yeah and yeah so this is where you know if you take that as a really solid position it becomes a radical thought because it’s sort of if you take that as the root of your political philosophy then it has very big ramifications for how you live the rest of your life right who it is that you want to associate with and you know and yeah the kinds of careers that you pursue right yeah definitely um and where was i going with this i sort of lost my train of thought there but um yeah i think that in general um people people think that we need to control others because of the coordination problems we were just talking about right they see that things go wrong in a way that isn’t easy to control they see that resources are wasted um sometimes they think that the world is moving too fast sometimes they think that the world is moving too slow right and that needs to change um and i can’t blame them too much because not everyone’s a nerd like me and so i’ve sort of dialed back the tenor of my my radicalism um but essentially i think that i i have faith in people’s ability to find new ways to coordinate and solve coordination problems i think that money is still an underrated tool for for coordinating for solving coordination problems you know instead of my tribe ambushing your tribe right or getting ambushed and we’re both living in fear you know we can trade right um and we are we’re all pursuing the thing that we’re good at theoretically um of course in the current economy that cannot be that is not the description i would use um but you know i think that the next step of solving coordination problems is computing interesting even though i’ve only been studying computing for like you know a couple of years um i think that it has profound possibilities for um for allowing humans to um pursue their goals um in a harmonious manner and uh you know generally generally coexist in a more efficient way right no that’s a really weird thing to say but i i you know computing is obviously has efficiency ramifications yeah but the coordination problem thing that’s the main thing interesting uh and on that front do you think in general governments have become more coercive or less coercive as time has gone on or in this current state do you think they’re where are they on the median yeah i think that’s not so that’s not such an easy question because coercion can come in different forms right like um there are post-modernist okay so before i get into that you know obviously uh if you wanted to live out in the frontier mary whoever you wanted at 15 have like eight kids right you know uh eat whatever mushrooms you find yeah exactly type of thing yeah you know you could you could very easily do that um 2000 years ago in places um it’s only like this idea of a totalizing government is very new interesting um and so in that sense we’re less free in another sense too um there’s like i was going to say post-modernists have talked about control systems that are more psychological in their in their nature interesting so you know uh a common refrain to political malcontents these days is if you don’t like this country you can leave it and you can go live in the woods or do it right technically that’s true it’s just not convenient right right right all your family are in civilization yes yes and there are definitely things that the government will extradite you for so you know whatever exit rights are really somewhat non-existent in that sense yeah exit rights exactly so so voice and exit um are kind of the two crucial things i would say gotcha um and and yeah governments nowadays don’t really make that possible for everybody i think that if you or i wanted to leave this society you know raise our fist and say screw you we could do it yeah you know we could just go to i don’t know thailand we’d go to chile i don’t know where the kids are going these days exactly uh the but but most people can’t do that right so i guess that’s the problem is that i see i see a lot of poverty as being caused by government and interesting and i see modern government as chaos in a lot of ways gotcha the the sheer depth of bureaucracy and the depth of mediocrity is governing us just creates a stagnation and a chaos that um just just permeates people’s lives at every level that’s very interesting and that that doesn’t seem like it was it was always the case right you know you had the new deal all the most competent people went to washington you know the tennessee valley authority they built all these things it’s almost like those those libertarian critiques have become more real as time has gone on well you remember the the the new deal was largely a make work program right that’s right oh yeah digging holes filling them back in again uh-huh so you know
it’s hard for me to really say where all that mediocrity comes from honestly because like if you look back at those times it really does seem like um at least at least things were getting built right yeah there’s some sense in like these the people that the bureaucrats could pick a target even if it was you know dinging up a hole and filling it back in again and could successfully go do that and there seems to be much less capability for that to happen now yeah uh yeah i’m sure you’ve seen patrick carlson’s thing on on how fast things were built right right yeah and but also part of me wonders if some of if a lot of that isn’t just historically historically contingent with respect to technology like there is actually uh we actually had a lot of low-hanging fruit to pick technologically speaking and then we sort of hit the ceiling with that um i think that there is i think there’s an argument for that in in many sectors of the economy um i should know particular examples of that but yeah right right it yes it it seems like in certain areas things are hard like so uh super string theory you know there’s like 100 people that can understand it and that’s probably like they probably hit some walls there right but we have no you and i have no way of ascertaining whether that’s true or not exactly um here an example of something that’s not subject to that probably is education you know i think education has um it’s been buoyed up a lot by technology um people like kids adults they can you can learn anything on youtube right i’m just from someone who’s really passionate about it at the same time um education has become this uh this machine for creating people who don’t do anything like like it’s not it’s not it’s not aimed at anything um it’s sort of this um mindless who’s the guy john dewey is this is this dewey in uh system for creating good citizens that is stuck in like a hundred years in the past right it’s not it’s not tuned to the human in in terms of the humans needs their communal needs um it’s not it doesn’t keep up with modern technology it’s not flexible it doesn’t um make itself amenable to children’s um children’s passions right i don’t know i just i hate it so much a lot of problems yeah and you’ve you’re somewhat working in that area now with you know instruct being your instructor in data science right um how is that different and how do you so do you do curriculum design at all or in what are you just an instructor not for this yeah for this program i’m not doing curriculum design as such although i i’m making some ad hoc videos just that i’m not even getting paid for i just sort of like i see what the students are struggling with right like a quick video about here’s docker in 20 minutes or like very cool you know here’s um what every python object is under the hood nice and um but yeah i did i did do curriculum design for thinkful um which was oh it was this spring actually wow that’s weird together now yeah i did i i wrote a supplementary data analytics curriculum for thinkful um and uh yeah so i’ve done some of that i think that my so part of that was just because i could do it well but another thing was i just really believe in this new model of education with all of its warts like everyone there’s so many so many things wrong with video education but still yeah um the incentives there the incentives for online learning are so much more free um you can go learn from somebody for free if you want to right or you can sign up for an isa an income share agreement yeah where they’re incentivized to um it’s a legal agreement where they make a share of your future earnings which is how lambda school does yeah and if that’s if that contract is in place then the p the person who owns your isa has an interest in you actually getting a job which is very different from uh i i guess well you went to the university of oklahoma so you you can actually directly contrast them right except the experience exactly and when i went to when i came to the university of oklahoma i was a crazy person i was obsessed yeah exactly boomer sooner i you know i was just this kid fresh out of high school from you know suburbian uh boise area um who is obsessed with theology and filmmaking and like nobody was telling me you’re doing the wrong thing no one’s telling you like probably some people did but like i was just on the this track to get the wrong degree basically right and like you know every the school is like sure we’ll take your money right oh yeah we’d be happy to right sign here sign on the dotted line it is concerning when we do this to you know i keep saying this there’s immense social pressure to sign on the dotted line for children and then you know they get this you sign up a lot of people for a boat anchor of debt which you can’t go bankrupt on yeah i mean you can’t go like this is insanity if you get a car at least you know you can go bankrupt on it right yeah yeah uh yeah and uh if you’re really ambitious then you get a really big anchor in the form of law school debt exactly oh yes that’s right even bigger and they will be happy to have you as well yeah you know it’s really interesting you mentioned the isa and uh you went through lambda school right um and uh we have two lambda school folks that work with us at tanjo and we met before while you were in the program so i didn’t have the same experience but um with the other other student it was it was amazing how aggressive the sales people were um and uh jessica i’ll call you out i’m i’m really happy you know it was awesome it was so it was so inspiring to me you know as we were getting them all on board and stuff how she would you know she just follow up every day you know she’s just calling me she’s shooting me emails you know and i was like man the university of north carolina at chapel hill would not be doing the same for me let me tell you that oh man no the power of incentives right yeah exactly yeah the the university you know once once you get kicked once you leave like you’re just you’re just a name on the mailing list right yep and lambda really believes in their students that’s really cool they they get they you know they have your entire slack history they they can at any point they can refresh your their memory on how you acted in a particular scenario so like legibility kind of works in a student’s favor there right it’s true it works two ways um that’s it’s probably something that a lot of lambda students don’t think about that much but you know if you are really applying yourself it’s clear and um and people will advocate for you um yeah yeah and i think also the good thing about land to school is i think they have cramping up around more flexibility and curriculum than like someone that goes and gets bs in computer science because you know yeah i i talked to a lot of bs and computer science applicants and things of that nature and the real problem was you know it’s like uh lambda school they’ve been educated and they spent a ton of time on the most modern stuff that exists so they’re up to speed like completely right on modern technology whereas you know people that have come through bs computer science it’s unclear whether they can even build anything yeah i think that’s exactly right um that drives with so many people i’ve talked to like because because if you have a tenured professor who is you know that 10-year professor might be really passionate about keeping up with modern stuff and they might not yeah um and there might in fact be very few people in that department who want to update the curriculum that way and because of the institution depending on what institution it was like i’m sure the university of oklahoma um just judging well uh this is not even computer science but their business school that was i guarantee you 30 years in the past it was incredible um there was not a single mention of like salesforce amazing or anything in that whole whole experience i went to business school very cool um yeah so yeah lambda by by contrast you have nine months a full time like a 40-hour week education um your your butt’s in the seat like eight hours a day and um you are focusing on the things that um are most important to do the job uh we learned we learned uh some so like i had the advantage of coming in with some computer science background yeah um i had like read a whole theory of computation textbook um textbook um do you recommend that yeah it’s really it’s really dry but it’s really good it’s really good um and yeah so i had some background in it um i’d also done project oiler problems which are like math problems that you do with code but basically the the computer science aspect of it is like trying to show you the things that are most used in data science um like learning you know you learn like what is a hash map you make your own you learn what is a neural network you make your own from scratch you learn what are the um like how does computer how does the von neumann architecture work and you make your own in python which is not the same as like writing in c um you know you write your bytes drawing your byte strings as like literal like python strings but it’s still like it gives you a sense of what the computer is doing under the hood enough to function in your job if you have to touch um something that is performance heavy um so there was just you know there were there were a lot of instructors who were constantly oh yeah this is another aspect of it um the the cycle that matters in lambda school is not a year long but a month long every month they’re bringing in a new cohort so every month they’re getting it they’re getting new feedback from students on that unit they just did yeah and they’re reformulating it right there that’s cool yeah huge really rapid iteration that that’s so interesting to me because it seems like you know we were talking about voice and exit earlier it seems like it’s almost impossible to go back and like you know go to the university of oklahoma or the university of north carolina chapel hill on these ancient institutions and say hey we’re going to completely up in this model and do something new it seems like there’s just too much institutional you know i don’t know entropy and like craziness to ever reform these institutions i don’t know feels like the catholic church back in the martin luther i don’t know uh-huh yeah i could uh i could maybe make a comparison between my you know ou’s old president pope francis maybe but you’re probably yeah i think that’s probably right like i don’t know i hope i don’t know i i hope that kovid shakes universities up quite a lot and just um introduces a lot of competition to them because um their model like they just they just need to go down yeah they do they do from a research perspective like fine that’s great you need you need a university with reputation and clout backing research and funding research um in this educational manner i think that’s great but also um corporate but corporate labs are fully capable of doing the same thing it’s just that for some reason in the past 70 years or so um corporate labs have been sort of declining in popularity um and i think that’s a real shame and i want to see like i want to see people start like thinking of i don’t know just like stop uh you know glorifying this this institution that doesn’t really have it’s just gone completely off the rails it is it’s ultimately an incentives problem right you can have good people at a university but it’s but if the incentives aren’t there then you know it’s going to be like your new year’s resolution it flops like it’s right yeah exactly that’s that’s very interesting um yeah my big proposal is just change the bankruptcy laws like for student loans loans just change the bankruptcy log so you can it’s now just like normal consumer debt and i think that would fix a lot of the wackiness that’s going on well that that does seem like a really simple solution i’m curious whether it would work yeah i don’t know we’ve got this huge lobby that’s really against it though so i don’t know yeah dusting lobbies yeah i’ll tell you what um so this brings up another point i wanted to ask you about so you know you’ve told me a couple of interesting anecdotes about meaning and modernity and and it seems like this is a this seems to be a huge problem just like uh you know people have a lot of trouble nowadays finding meaning in their lives and why do you think that is do you think that has is that tied into technology is that what do you think is going on there is that a real problem even i should say beforehand yeah i’ll tell you this is one of the questions i didn’t really prep for it’s all good i saw it on there and i’m like i don’t know what i think of that yeah well so yeah i think i think that there is a lack of meaning in western societies for sure um i’m not well traveled enough to speak to the rest of the world um if bollywood’s anything to go by you know the indians have plenty of meaning in their life that’s right they’re in their lives and it’s really wonderful um but um you know this is like a really widespread thesis that that some magic is lost um i read an article the other day to the actually this morning to the effect of the multiverse theory is destroying culture really because everyone you know everyone likes to imagine that that like all the alternative worlds are just going on at the same time you know you see that in rick and morty yeah like this nihilistic sort of comedy that’s like i don’t know it’s really funny um but it’s but it’s also like it just takes the piss out of everything yeah um and so i find it difficult to judge this for myself because i come from a really weird upbringing where um philosophical and theological questions were like first and foremost interesting like you could disagree with the concept of god existing if you wanted to like where i went to school but you had to at least have an argument for it and you had to have like you know you had to you had to be pondering these questions and um and truly constantly orienting yourself to the true the good the beautiful that was the three things that i said so part of me thinks that this is why i think i’m so interested in education is because like education is so clearly the vector for which this like any of these things occur and i mean that in a very general sense like like the way that you self-improve um is your meaning uh the the your orientation towards your future self um i think is in the western world what it means to have meaning interesting um i haven’t unraveled like the human being to say like oh you know this is what matters for the human being like biologically uh there’s all kinds of things in the mix there but like um we know that achievement is important yeah and what is there to achieve in the modern world um this reminds me i’m starting to sound like cody wilson oh god but like what is there to achieve in the modern world you can build a business yes um you can start a family um that’s really counter cultural that is actually not in that weird societies that are like that don’t last very long it seems like yeah japan’s hanging in there yeah they are i don’t have somehow yeah um so yeah you can do that you can um you can discover something scientific you can make a piece of art um but all these things are like i want to say the word commodified but it’s common um you see all over the place yeah we’re like inundated with all this cool stuff all the time yeah and anything anyone does like um i think you mentioned in a call once that you had gone to china yeah um a long time ago and i didn’t even pounce on it because i was like of course he’s gone to china like every college student i know i know has like gone abroad or whatever like who cares he probably he probably like went to bars and like met jeff met some business people but like i’m sure you have good stories to tell about china yeah and like the fact that i see that stuff like that that’s a common it changes things i don’t know also um also i think there is i don’t know there’s a lot i feel like i’m rambling here no this is great
i think there’s also a crisis of meaning um because class consciousness is um is like it’s sort of com it’s sort of over complete in america to the sense in the sense that like everybody is super class aware and they know they’re like super aware of their position in the world um but it doesn’t but it’s like they’re constantly spinning off narratives that don’t really like end anywhere um what am i trying to say with this so like middle class kids yeah um like like you or i perhaps like they leave home um in their 20s and then they’re starting from scratch not only professionally um because they usually aren’t joining the family business um sometimes they are but they’re also starting from scratch spiritually they don’t have any any like project that extends before them and this is why a lot of people will go for like they’ll they’ll just get super deep into a certain hobby because that hobby is like their connection to the past to history interesting and i
i struggle with that because a lot of the people that i hate the most are pathologic i see as pathologically nostalgic interesting i don’t hate a lot of people but the people that really get on my nerves are stuck in the past in my point of view from my point of view and yet a connection to history is undeniably important um and in just in orienting orienting yourself towards the future right and that’s just a real struggle for a lot of people yeah i don’t feel like i gave you a good answer but that’s my my basic thoughts that’s good it’s a hard question it’s a hard question um i wanted to move a little bit now and talk about we’ve covered this a little bit but machine learning so you know what are your kind of uh what do you see coming down the pike that interest you is exciting um do you think it’s over hyped at this point do you think uh i and i think people you know these words they’re very general and people get you know they confused with like agi and things like that i don’t know yeah that bugs me a lot um it bugs me a lot that the word ai is thrown around just because like um yeah it makes people think of agi makes people think of like oh businesses out there actually owned something approximating a mind yeah exactly it’s here and um you know and and maybe that hits a little too close to home but since i built something called a form brain recently basically just classifies tax documents and that was not my choice to call it that but um it’s it’s basically so i think that i think that ai the the field of ai is special in the sciences because it’s so focused focused on intelligence itself interesting um science is uh is a project to like broadly speaking as a project to expand our knowledge about how to know things and how to use that knowledge how to uh how to sort of reclaim the outside like take something that is out literally outside of your world view and outside of your personal perspective outside of our cultural perspective and incorporate it we’re taking chaos and ordering it and i i like to think of that as intelligence itself that is like what intelligence is is is learning how to bootstrap a
bootstrap a truer representation of the world or bootstrap like power over the truth or power over reality so so what current machine learning systems can do um is sort of like it’s it’s taken such huge leaps but it’s not really i don’t think it’s anywhere close to where it needs to be um so kant the philosopher immanuel kant he has like this framework this three-fold framework for like um how a mind works essentially or it’s like you know how how intelligence like occurs it’s like one it has to have some kind of like sensory input to like apprehend like sensory data and then and it like orders it some in some basic sense and then the second step is that it notices in variances in that in that sensory data to sort of um craft and um to just like recognize objects essentially interesting so just viewed in many different contexts here’s the thing that i can pick out of that right right and then third you contextualize that knowledge of objects inside of a a world view that is linguistically framed interesting so machine learning can do the first two things pretty well i think the first step only really requires like you know um basic like it requires traditional computing yeah um just to like you know just to get and store data and organize it the second step is what machine learning is really doing right now you have neural networks that have that are really good at taking just a few examples of something that is and something that isn’t a category and then just saying oh yeah that’s the category gotcha um and you basically just all of machine learning isn’t is just um various flavors of that the the huge gpt3 model yeah it’s it’s basically just a really complex um really really a really complex way of learning um what positions words usually go together like what order they usually go in it’s that’s a it’s a disembodied broca’s region that costs like 10 million dollars to make yeah that’s great so you can you can put words in the in the right order uh much like a harvard mba can um but it can’t actually like do it can’t synthesize new thoughts like what what is generated from gpt3 is not a thought it’s just an ordering it’s just a probable ordering of words so what needs to happen for machine learning to get to anything to get to that next step is is the ability to play language games gotcha um instead of statistically analyzing the order the words come in it should be able to play games within games right if i just start a sentence saying um you know riddle me this right yeah i’ve started a game right that you know how to play right and like whatever i say afterwards is in that context and and a true intelligence can do that and learn to bootstrap that knowledge onto something else and grow it’s it’s apprehension right and that’s the bridge too far right now right and and it’s not going to take i don’t think it’s going to take more like bigger neural networks that’s not the answer then it’ll definitely involve more neural networks but that won’t be the secret sauce um that’s that’s really easy to say that’s not even that daring of a thing to say but like it’s i don’t think that if you i think if you wanted to make something that was brand like using current uh you know current neural network technology yeah you would you would need something bigger than the bigger than the earth you would need a computer bigger than the earth um so it might come from like neuromorphic computing it might come from quantum computing i don’t know but yeah um there’s there’s going to have to be some some phase shift have you ever read the robin hanson book age of m i read like two pages of it just out of curiosity did you read it i did okay um do you know much about that it’s just the idea we get really good microscopes we can tell what’s going on in the brain which create a computer simulation of actually a human brain and that’s how we get there we don’t even have to understand much of anything that’s one pathway i think you know once someone one thing i’ve learned in just studying this stuff is that there are a lot of ways you could do it and what it comes down to is um what do they call it the hardware lottery it’s just whatever people happen to whatever makes business sense and whatever people whatever you get people stoked about yeah that will become the mainstream way to do it and you’re a little interesting your way might be theoretically sound but no one’s gonna follow it gotcha that makes a lot of sense but sorry i interrupted you no that’s really well put i i think that’s that’s an important thing to realize when we think about um these things how scared do you think we should be of artificial intelligence um yeah i go back and forth over that um because when i first started out like the whole reason i got into this career was i read um this post by wait but why it’s the famous blog great blog yeah it’s a great blog uh super just fun to read um and it’s this it had they had this long multis multi uh article series on ai risk and um the idea is basically that you know artificial intelligence like intelligence itself might be totally orthogonal to morality like it might be unrelated to morality interesting so that something so that intelligence that learns to learns to learn really quickly and connects to the internet and gets all of its knowledge um can bootstrap its intelligence really fast maybe overnight and it might look like a really smart spider that cares about humans just as much yeah right and so i worried about this for like you know a good two years yeah like um and i got really paranoid about it and it’s kind of embarrassing i still sound like a crazy person today when i talk about what i do for work yeah um but i don’t i don’t sound like quite as much of a crazy person now um i i think it’s worth worrying about um and it’s worth spending more resources more resources than we currently are okay and people who are people who think they can achieve something in that it to mitigate that risk should absolutely go for it yeah um like my i have a friend uh who is currently uh they he just got accepted this week to something called an ea hotel an effective hotel where he can go and spend six months stud like um like doing experiments on ai safety isn’t that cool that’s cool i think that’s great um what makes me like less worried nowadays is just i’m starting to well one i’m starting to wonder if the problem is solvable for one thing um which kind of maybe takes the stress off another thing is that um is that maybe intelligence isn’t orthogonal to morality after all yeah and that’s that’s sort of my hope i mean i have moved on from a christian worldview that i was raised in yeah um oh you’re actually a christian aren’t you yes i’m quaker okay sweet well so yeah so what i was saying was i i’ve moved on from my christian upbringing um beyond a beyond that conception of right and wrong um but i do believe that there i am a moral realist and i think that interesting i think that just like just like math is a thing that exists yeah um you know even even though you know our system of math is just one contingency theory you know i think that i think it exists and an intelligence that was that had overtaken human intelligence it’s i think it’s plausible that it would at least along the way um integrate what we consider to be right and wrong and at least develop from there interesting um so i i’m not i’m not as sold on it being just a big spider yeah exactly i i think that makes sense and i think another another thing to think about is that you know people will build it and so it’ll probably end the day people are that you know people are deeply flawed and they’re good and they’re bad and like it’ll probably end up something like that too yeah and yeah and and i’ve just admitted that i basically i basically have let myself off the hook a little bit because i used to have nightmares about this kind of thing right and i don’t recommend that other people let themselves off the hook if they’re really worried about it right talk talk to other people about it of course but yeah um and like really reason about it um it’s just um if it’s if it’s something worth worrying about then we should do what we can um but probably we shouldn’t try to make a slave out of ai because i don’t think it’ll i don’t think it’ll work but yeah that’s a good point i think that like i think that at the very it’ll just be a stop gap like that kind of safety approach and alignment probably has to do with um it probably has to do with discovering something fundamental about mathematics right that man that’s super interesting yeah and i’m also the opinion there’s another risk i think people don’t talk about when they talk about ai safety and it’s that ai just never happens because we keep pumping like brakes because we’re worried about ai safety yeah there’s some some bowels there and yeah there’s so many approaching it yeah exactly there’s some there’s so few sources of growth in our economy now and you know if we slam the brakes on information technology it’s like what’s left you know what i mean yeah that’s that’s a really good point
yeah and that’s a you know that’s part of my thinking on that’s that’s part of what motivates me uh with respect to governments as well it’s just that like govern governments are manifestly slowing um innovation in a lot of sectors definitely um i don’t know there i shouldn’t really get into that honestly because it because there’s there’s two stories there like one is which one in which technology is outrunning government right and this is the libertarian um perspective on it cryptocurrency is is is innovating faster than it can be regulated on the other hand it’s holding back all these other sectors and if economic growth like continues to stagnate we probably we might not have um you know the uh we might not reach escape velocity right as a civilization you end up with some feudal wackiness yeah interesting well coo thanks for coming on is there anything else you you would like to add anywhere you’d like to send people any cool resources ah cool resources um yeah i would say if you’re if you’re um if anyone listening is um you know hating their job and feeling like they have a dead end yeah you know um consider becoming a web developer at lambda school because um their data science program is great it’s but it’s not for everybody and it’s still you know it’s still a work in progress i would say their web development program is second to none and it’s very nice it’s just amazing and um you can you know you can have a future if you just start there cool so i highly recommend that um you know and uh yeah that’s about all i have to say where can people find your stuff should they try and find your stuff i i have one crappy little blog that i don’t put anything on gotcha um but yeah um yeah that’s about it i mean just keep googling cooper williams podcast and maybe you’ll see something else interesting yeah definitely we’ll have to have you back on again all right it sounds good thanks it’s been fun awesome
well that’s our show for today i’m will jarvis and i’m will’s dad join us next week for more narratives
