Salon Dinner on the Future of A.I. Infrastructure

hello welcome to today’s talk onexpanding access to public a models myname is Brandon Jackson and today I amvery excited to share the work that I’vebeen doing with folks over at theemerging Technologies team at ASPdigital about public AI now in case youhaven’t heard public AI is um a movementthat is uh coming from all corn ERS ofcivil society andTechnology uh to figure out how do webuild infrastructure for the common goodso you can check out our white paper atthe link below um as well as uh join ourNetwork we have a slack Channel ifyou’re interested and public AI looks atthe question of AI development forpublic good a little bit differently andfocuses on infrastructure as the key topublicbenefit as part of this work we H cameup with this definition that has threeessential requirements for public AIwhich draw on the long history of publicinfrastructure and these are PublicAccess public infrastructure shouldmeaningfully expand access tocapabilities leveling the playing fieldpublic infrastructure has to beaccountable so that the public has someform of ultimate control and that it cantack an independent course in the rightdirection and lastly public AI has tocreate permanent public goods that aresustainably funded and maintainedpreventing private enclosurenow we didn’t come up with these thingsrandomly we came at it by looking atthis long history of infrastructure andtoday I’m very excited to dive deeperinto that history and specificallyunpacking what does it mean whenever youdon’t just have utilities that deliverthings uh deliver Commodities butactually when you really commit to theidea of expandingaccess so to kick us off we’re going tobe looking at lessons from this historyof publicinfrastructure now we all know and lovepublic infrastructure it expands accessto Commodities that we all depend onlike water power Transportation newsbooks andHealthcare and I like to think of it asactually a superet uh or a close cousinuh of any other firm so utility firm uhsells products to customers um and thenthey take the revenue from the productto sell and they reinvest it to buyinputs they need to deliver thoseproducts so if you’re selling water wellyou can’t give it away for free becauseyou got to pay for electricity laborcost and Machinery right totallyuncontroversial private company publiccompany the basics are the same and thiswhen you bring this product to Market itcreates value that people are willing topay for call this marketvalue drinking love paying for things todrink hygiene love to stay clean totallywould pay for that industrial uses wellpersonally don’t have any but I know alot of people do and I use the stuffthey make so all of these things areforms of market value and this littlemodel here is uh great and and it’s it’suh the model that you know most privatefirms follow they’re primarily concernedwith what is the market value of theproduct they’re making um and how can Imaximize my margin while paying for theinputs now here’s where things getinteresting there are certain thingsthat are actually more than just theirmarket value hard to believe Iknow these things are said to havenon-market or social value and whatwe’ve done over time if you look in thelong day view is actually we’ve iteratedour institutions to create waterutilities like water firms that actuallydon’t just look after your market valuethey’re actually designed to preventmarket failures and make sure theseuniforms of social goods are alsoproduced so if you at public watercompanies they’re they’re all about theindustrial use and they’re essential forthe economy but also they’re essentialfor nature they’re essential for publichealth and they’re essential forrecreation and what we see is that overtime a lot of these institutions havedeveloped really nice and healthy waysof of setting the price of the commoditythey sell at the true price that looksafter the market and social goods andthey do this by taking revenue andreinvest it not just in the marketinputs but going further and reinvestingin the quality controls and Commons thatthey depend on investing in for examplethe maintenance of the river Network themaintenance of the pipe Network and themaintenance of the sewage Network aswe’ll see you don’t have to do this butthey do it because and that’s what makesthem public they go above and beyond andI think it’s a really helpful frame forthinking about what what are thesepublic infrastructure firms doingbecause we can clearly slice out whatare the things that we get would getanyways from a private firm but what arethe things that go above and beyond andwe can look at Power for example so youknow what we get for free in the fromthe market um is the ability to havelife at home I’m sure electricity wouldstill exist um we have Commerceelectricity you know there’s a realMarket need for it it’s not going tototally disappear without public actorsindustry not going to disappear butthere are these other things that mightum for example nature um is verydependent on the way we we generatepower um and for things like climatechange for example but also PublicSafety like you know there are plenty ofexamples through history of utilityprovision going wrong and people gettinghurt from power plants or you know allsorts of um other things like that uhand Ral and pollution and then RuralLife uh is another thing that’simportant if you just had privateutility provision there’s no guaranteethat everyone in the country would getcoverage but with public utility we werecognize there’s something specialworth protecting about real life and wereinvest and we all pay for that notthrough taxes but through the way thatwe pay for power and it’s sold at thistrue price that then is reinvested backinto maintaining not just not justpaying for Market inputs like powerplants Fuel and labor but also payingfor the dependencies and their upkeeplike air power grid and fuelsupply so that’s the good news and thebad news is that it doesn’t always goaccording to plan and sometimes privatefirms don’t reinvest to look after theupstream and downstream comments thattheir work um both draws from and feedsback into and when this happens thingsget dirty and sick pretty quickliterally for example if you look at theUK uh it has engaged in a why theyregard why they regard to be failedexperiment in private utility provisionuh in water and as a result uh Tim’swater has been underinvestigateddisasterous situation and it means thatpeople are now checking live dashboardsto see whether it’s safe to go swimmingat their favorite local Beauty Spot thisis a sickening State of Affairs to saythe least and is really a product uh itshows you what goes wrong when theseflows of money um are notreinvested now what’s interesting to meand I think will be particularly I hopehumorous to Future historians who lookback at this is that the next slidemight be seen as a bit of a leap but Ihope you agree it’s actually not andthat is a say that River ecosystemsinformation ecosystems aren’t actuallythat different in important ways weviscerally feel bad whenever watersprays in our face and we get a coliwhen we got going out for recreation andthat makes us angry but I would arguethat’s you know I’ll leave it tohistorians and say but you know perhapsnot actually that different than whathappens with our information ecosystemsit’s just we don’t have a tradition ofthinking of these things in the same waybut the good news is that we actually dohave a large scale natural experimentrunning right now we have countries thathave uh public utility provision formedia and countries that don’t thatdepend entirely on on private providersso if you look at the compare forexample the US and the UK you can see areally dramatic difference in the US uhthere are no State media providers andinstead most people get their media frommajor television channels Instagram TikTok or X these are private utilitycompanies that are providing a medcreating a mediaecosystem and they are not reinvestingin society and getting is milking it forall it’s worth getting us all addictedto feeds and you know arguably in thelong run probably making us a bit sickerthan a bit of eoli would from temp’swater but then in contrast with that notto say the BBC is perfect in fact it’shated by all political sides equallywhich itself probably a very good signBBC recognizes that if you’re going tohave a strong culture publicaccountability and cohesion you have toreinvest in connecting the citizens andbuilding content that reaches everyoneand not just the people who have themost money the result is somethingdistinctly British and it I can speakfrom my personal experience having livedhere during lockdown um going throughthe pandemic with these two differentMed utility ecosystems was completelydifferent here and I’m so thankful tohave been here and been a part of theBBC uh ecosystem here um but I thinkthat this goes to showthat this template of thinking ofutilities as an as a feedback Loop wherewe pay for services in order to maintaina healthy Commons is something thatwhile it sounds crazy on its face it’sactually something we’ve been doing fora long time and is actually one of theleast controversial aspects for oursociety unfortunately AI Foundationmodels are not being built in this waywe really need to be building them witha longterm in mind it’s not hard toargue that the AI baked into everyeconomic interaction in the F likelybaked in every economic educational andsocial interaction in the future shouldprobably be based on the equivalent ofpristine Alpine Spring Water rather thanPooh and the temps however we’re notexactly setting ourselves up for Successhere we have two dominant models ofproducing these uh these systems on theone hand you have private AILabs similar to the example with tempwater these institutions areincentivized to produce maximize rentand not to reinvest you already see thatfirms are starting to slash ethicspending and instead of accumulatinginstead of funding institutions who arepublicly responsible for creating newdata sources they’re creating their ownprivate data sources um and then payingpeople off in in developing countriesrather than contributing to the Eknowledge ecosystem um you know as aresult we might be facing seriousexistential risk um to say nothing of aweaken of C weakening of cultureknowledge production andmisinformation unfortunatelypublic-minded uh approaches thesituation are perhaps no not much betterthese this work often draws from thelong tradition of government grants toScientific institutions with the beliefthat if you just create knowledge thenonce it’s produced it spreads like apublic good and needs no maintenance butI think that as we’ve seen maintainingthese comments can be expensive and it’soften a price that we collectivelydecided is worth paying but what’s soremarkable to me um from looking at thisis how almost none of the examples thatwe have of public institutions uh raisemoney from what they do and and in orderto fund sustainability and instead theygive it away for free so without thisthey’re left facing not only uh aninability to reinvest in maintenancesecurity and safety uh but also not evenenough money to properly cover theirinput cost depending on state subsidyfor everything it’s not a sustainablesituation and it’s setting ourselves upfor a collapse of thecommon I argue that’s why we might needto start thinking about things a littledifferently and maybe we should by say Isay we should think about thingsdifferently let’s try thinking aboutthem the same way we’ve been thinkingabout utilities the whole time which isthat we can have utilities for AI thatsell AI at Cost but not just at a raiseto the bottom price we can sell it atthe true price and because we’re talkingabout low numbers here most firms willprobably be happy to pay that just likefirms are happy to pay whatever thegovernment regulator says they need topay for water so what might happen wellas we’ll see some special things happenit’s not just a matter of financialstability it creates a whole newincentive rure public utilities arerespons are on the hook for drivingusage because their revenue andoperating budget depends on it thismeans that they’re incentivized to dothe cultural work to drive adoptionacrossSociety this Revenue could not just fundbreaking even this Revenue could fundactive investment uh into growing tocover to create new forms of socialvalue and to invest in the resiliencethat we need to ensure these systems aresafe and dependable just like tap wateruh is in our our dailylife sonow I’ve been talking about flowchartsand while I’ve been mentioning somehistory um I’ve I am really excited nowto really dive into some of the lessonsfrom the history of Public Utilities soas we’ve seen from the flowcharts publicinfrastructure doesn’t just do what themarket does and so it often does thesame things the market does and thengoes above and beyond and that’s whatwe’re going to be talking about what isthat Delta something in our definitionwe talk about when we say we want toexpand Public Access or level theplaying field these might sound likeabstract things but the good news isthat there’s lots of fantastic examplesthroughout history of Institutions doingthis so we’re I’ll uh go over them nowand we’ll walk through one by one sothese institutions have amazingincentives to invest in adoption to goThe Last Mile serving everyone to focusR&D on adding value at the point of useand to reinvest for resilience so let’sgo through these one by one so let’sstart with thinking about um the goal ofwe might already have for AI we wouldlove the distribution of benefits of AIto bewidespread but we’re on track for a bigmarket failure here because on the onehand you have a resistance to automationon the part of the public which meansthat usage of these tools and a sense ofpublic the public having a stake in theconversation is very low it’s a bit of agim but people don’t really feel likeit’s theirs right and instead the onlypeople who are really driving homeadoption are firms because they aredirectly incentivized to replace laborthis creates a really crazy landscapefor AI Builders because even if theywanted to make money from the public thepublic isn’t that interested and if youwant to make money from AI at this pointAutomation and replacing humans as fastas possible is the only game in town andwe can see that this process alreadyhappening with over 80% of this year’sYC demo day entrance AI apps that areautomating things uh that I think wouldhave blown people’s minds many years agowell what’s the alternative to this wecan have public institutions that areactually incentivized to drive adoptionand incentivize making a technology apublic concern getting everybodyinvolved in theconversation there are some greathistorical examples from this from herein the UK um my favorite example andapologist if you ever heard me talk youwill have heard me talk about this isthe BBC’s licens feew model this afascinating anecdote from the history oftechnology that not many people realizethe BBC was actually funded founded inthe beginning as a Consortium of sixelectron manufacturers that’s right itwas the British Broadcasting Company notthe British Broadcasting Corporationthat change happened later in thebeginning they were set up they weregiven a monopoly Charter by the postoffice there was State involvement butthe institution of the BBC wascompletely on the hook to drive radiosales which is the incentive of the theelectronics companies obviously becausetheir license fee was charged at thepoint of the radio sale not as an annualusage fee this incentive alignment meantthat the the BBC needed to reallyquickly understand what were thebarriers that were preventing peoplefrom bringing radio into their homes andbuild the content that would make themwant to go to overcome those barriersthis then kind of primed the pump if youwill and led to an explosion in radiodesign that brought over the shortperiod from 1922 to 1934 transformradios from something that looked like ascientific apparatus to the kind ofbeautiful plastic modern objects thatwe’re we now associate with the erasomething like you’ve seen a porroepisode that transformation happenedvery quickly and it’s because there wasa public institution that wasincentivized to overcome these barriersand was directly on the hook and itturns out that the BBC still is on thehook and I can I can report that anytimethe BBC’s usage numbers drop there is astory about it in the newspapersespecially the right-wing newspapersthey love to hate to be fair they hateon everybody but especially the BBC andthese numbers are broken down bydemographic and you will you will see inCrazy articles say things like the BBCis racing to figure out its problem likewith under 35 minority groups andbecause they know that that’s a segmentwhere us has dropped and the BBC cannotignore this it’s not like governmentthat can just be like yeah whatever theywill literally lose money and staff havelost their jobs as a result of thisalready and I think that kind of directaccountability is fantastic it’s atotally different way of operating thanthe way we think about government whichgets a steady tax income no matter whatit does these institutions are oftenincentivized to drive usage and to becreative that’s what we need anotherexample is underground uh when it wascreated it was something that was thethe idea of taking a train in the cityalthough it seems common to us was notsomething people did uh you have toremember electric electric lighting andMass public lighting was a relativelynew phenomenon in the 1930s and peoplejust simply didn’t have these Traditionsor habits and as a result the t uh wellthe early prototypes of Transport forLondon invested heavily in advertisingcampaigns because they knew they hadextra capacity evenings they were likehey go to Brightest London and come homeby the underground they created new theyidentified the things that would the usecase that the public would want and thenthey promoted it the same thing forCountry walks this is especially uhpopular during the inter War years whenthere wasn’t much money going aroundthere’s a big promotion of hey why don’tI just go for a Country Walk uh becausethey had extra but it was this happyUnion where on the one hand you had apublic supplier who was incentivized todrive usage but to do so in a way thatwas good for the public so they whenthey found use cases that were win-winThey promoted it like crazy buildingfamiliarity and now being leading toother forms of cultural infrastructurewe still have in the UK like websiteslike the Saturday Walkers Club where youcan go and see a crowdsource map of allthe places you can go by train on aSaturday for a walk uh stopping at a t-room and ending at a pub all thosethings got their start from this exactincentive alignment that we’re talkingabout here having institutions that arebecause because they charge money areincentivized to go and bring the publicalong on thejourney so that’s something we can dofor a public a we can build institutionsthat depend on usage and we canrecognize that this cultural work ofbringing of identifying use cases andmarketing it is actually part of theprocess of building publicinfrastructure it’s a win-win and itmakes sure that we all have a stake inthis rather than Tech remaining the thetech like AI remaining just somethingfor the elite who are trying to automatejobs as quickly aspossible next up we ideally want AIworks for everyone but right now we’realready in a world where we have bigmarket failures where AI unders servesmany cultures and many languages andworks really great if you’re um anEnglish speaker from San Francisco andthe further away you get from that Normthe more things go off the rails so thelesson from history is that publicinfrastructure can go take money fromeveryone and use it to reinvest to gothe last mile to service everyone nowthis phrase the last mile is veryphysical and the the metaphor um doesrequire a bit of imagination herebecause the situation is a bit differentAI but actually you know I think thehistory of rural axis is actually veryinspiring almost every us utility hasinvested heavily in rural access umpaying for things like power lines toNowhere by nowhere I mean small villagesrural Farmers paying for roads to placesthat currently don’t have anything therebut believing in the possibility andlaying water pipes um to to ensureeveryone has access to clean drinkingwater so the revenue generated from theusage this to the more urban areas hasbeen quietly redistributed to pay forthese things often with State subsidiesbut um it’s had a big impact I think thebest thing to show showcase the kind ofmetaphysical societal impact of thesekinds of Last Mile initiatives comesactually from the Postal Service when inthe middle of the 19th century thePostal Service adopted a flat pricingscheme so it was the same price to sendposts from one side of the country tothe other this was an incredible uhmoment and it led to a45x increase in Postal Serviceusage and what it did more than that itdidn’t and you can imagine all the goodsand services and just so much so muchmarket value wow so much value okay butit also did something bigger than thatand it changed the collective identityof the country like we’re talking aboutwith the adoption it turned this thingfrom a niche concern to a public concernit made everyone a stakeholder andthat’s something and it made everyonefeel like they’re part of somethingbigger and that if they could take riskthat the poity would back them up um andsupport them going a little bit outsideuh leaving behind the big city and goingout and seeking opportunities elsewherethat had a big impact on people um andsomething really special happenswhenever you have these institutions andin utilities that are reachingum every place um and the example I lovehere is with libraries wheneverlibraries open in a community that meansthere’s a public servant dedicated tothat Community who’s able to listen andlearn and figure out what’s needed andsurface learning surface opportunitiesback to home base in a way that justwouldn’t be possible if you only hadlibraries in the big cities and I thinkthat they often you know would thinkingof examples here in the UK um the WhiteChapel library was famous uh for havinga massive multilingual uh uh selectionbecause it was the home in the East Endas it was of a large Jewish Communitythen a Bengali community and it hadmassive libraries and all theselanguages that everybody had what theyneeded and I think that this is exactlythe kind of investment that made theEast End such a special and welcomingplace that gave us all sorts of thingslike for example the rise of theanti-fascist movement was started um inthe battle for cable Street which isjust down the road from the white chelLibrary where all those communitiesbanded together to fight off the Marchof the fascist into the neighborhood in1930s that’s just a random example Imean but they’re they’re kind of inlessexamples of whenever you have publicinstitutions that create space foreveryone people start acting andbehaving differently it’s a reallyhealthy way for communities to stay uhconnected so I think when we think aboutthis from an AI perspective we need toreally understand that um you know whenwe’re thinking about should we pay extrato make sure that everybody is connectedum there are lots of historical examplesthat show both the economic socialpolitical consequences of that be verypositive and also extremely unexpectedum and I think that that is somethingthat you know is currently missingwhenever academic research institutionsare building AI from the Ivory Towerwhat they’re missing is that word fromthe front lines that’s why it’s soexciting to be uh working with Folks atLibrary of Congress I know metov islooking at doing a project of gettingpublic AI workstations into librar startfiguring how do we learn from what’sgoing out there do this research toidentify these needs and feed it betterthat’s something that great going thatlast mile is something that publicinstitutions are truly world classat so third opportunity here is to scaleaccess to key benefits so let’s imaginetheAI does everything that they say itmight do it it not only cures cancer butit’s like expands access to Legal careamazing per personalized education rightso let’s just say for secone that existswho’s going to benefit fromthat so if you look at the history ofInternet you don’t need to look too farto see that a lot of those things havecome true but for whom they they mightbe even if the fundamental researchhappens inside of public institutionsthey’re productized by middlemen whotake world class research bottle it upand limit access for a profit and thisis something that public institutionpublic infrastructure and publicutilities are actually really wellplaced to do and the lesson here is fromthis history is instead of doing thebasic research you can actually Identifysome really critical features and buildthem into the existing networks thateverybody gets access to them um and Ithink there’s no better example of thisthan Bell Labs um so Bell Labs uh had avery uh capitalist beginning of coursewhenever um I was under Grand Bellreceived a patent for the telephonecreating a spinning into existence anincredible commercial Monopoly but overtime babbs became increasingly regulatedto the point by by the end it waseffectively a public utility um they haduh things like profit caps and andthings of this nature are reallyinteresting to look into the way it wassteered in and and brought under controlprogressivelyumso um so what did they do differentlythey brought they created a place astoryed institution called Bell labs andit was in a sleepy area of New Jerseyand they brought in the world’s bestscientists and put them in rooms crossfunctional teams as we call in uh intoday’s lingo with engineers andtelephone experts and what this did wasit created an ecosystem where the worldSP scientists weren’t just doingfundamental research they were doing thefundamental research that had a directpath to improving customers experienceat telephone Network so instead of justgetting breakthroughs in physics whichthey did you got transistors whichenabled longdistance phone callsand you got automation that enableddirectdialing and and you got DSL which couldbe gave people highpe internet accessthrough existing networks these areincredible breakthroughs but they’reones that didn’t require any middlemenand in fact could be delivered directlyto customers and I think this is sointeresting because AI is exactly thesame technology as telephone line if youcan establish a trajectory of having apublic Foundation model anytime a newtechnology that is accessible via textcomes online it can be plugged into thispublic infrastructure just like DSL wason the existing telephone lines but ifwe don’t have this public institution wecan fully expect that extra feature tobe charged as premium add-on and to onlybe accessible to those who can pay youmight think of it as like you’re it’slike if as if the highways are gettingbetter and better but only the rich canafford to use them that’s kind of aworld that we want to avoid um anotherexample of this is uh I think I’mlegally required now as a British Citito say our NHS the National HealthService here in the UK um and and what’sunique about the NHS is that uh it is apublic healthare a vertically integratedpublic healthare institution um and oneof his foundy principles what it iswhich is something I heard again andagain when I came here and I was a bitmystified by is that it’s free at thepoint of use and what that means is thatwhen you need it there’s no money youhave to pay at that moment there are nomiddlemen there are no hidden fees andthat means that you can be fullyconfident that it’s yours and thatyou’re going to be taken care ofthis incentive to basically do endtoendservice is super interesting becauseit’s created this incentive to have tobuild up an Institutional process foridentifying opportunities and forcreating good value for money service sowe have in the UK the National Institutefor care Excellence called nice and thisis literally empowered to make thedecisions about cost benefit ratios sothat everyone has access to the maximalgood for the best bank for Buck if itfinds a good M for moneytreatment just like DSL was broughtonline through the telephone Network theNHS can immediately bring online a newtreatment and scale direct universalaccess to everyone this is amazing andabsolutely a model we should be thinkingabout when you think about the long-termtrajectory of public AI so myrecommendations here would be to make tonot just fund AI research butspecifically set up these places whereyou can have a pipeline connectingdirect re where breakthroughs areconnected directly uh into the servicemuch like DSL was and long distancecalls foring telephone Network this willcut out middleman and scale access topublic interest Technologies now ofcourse there’s a judgment call to bemade about what kinds of breakthroughsare best scaled through the privatesector um I personally am a big fan ofthe NHS model but um you know just to ERother other side of the argument youmight say that certain things likegeneric um generics that should bescaled should be scaled acces genericdrugs but that we actually need a wayfor to like through the patent system toprotect the rights of private firms toincentivize them to make thosebreakthroughs that’s a totally fairargument the question is to me um likefor the things that are are made in thepublic interest how do we scale accessto them and just create this expectationthat you’re welcome to create a businessthat’s a fantastic idea but that you’renot going to be able to create abusiness just from being a middleman tosome public research we want all thegreat public research going there to bea lowcostoption for it um so that and then thatcan focus business effort on eliminatingthe remaining barriers B for is justhaving the technology there isn’t goingto solve everything people aren’t goingto know how to use a technology theymight want need supplementary Servicesthey might need the the the system mightrecommend a product and they need tofind it these are all things that are afantastic way for the business sector tocontribute but um at the same time whilenot getting in the way of these corevalues at the point of uselastly um AI could strengthen Societybut as we’ve already seen from theexample of water companies there’s areal risk of depletion of the Upstreamecosystem um and pollution of thedownstream information ecosystem sothat’s why utility companies and AIcompanies or utilities should reinvestinresilience um so the example I’d like touse here as we’ve already talked aboutwith the water companies um is themonitoring and protecting of wholeWatershed regions um andactually a lot of this infrastructurewas built in a crisis moment and I thinkit’s really interesting this this greatstink of 1858 because the the Tims justwas extremely smelly and rotting and andthis was very noticed by Parliamentwhich is based right on the side of Rivtemps so this very visceral experiencecalled the Great stink um actuallygalvanized Parliament to drive changeinto build public infrastructurebuilding some incredible Waterworks umthat were talentedin you know as one of the great wondersof the age at the time and you know youhave to wonder what is the great stinkof AI going to be you know I I don’tthink we know yet I think we’ve uh youknow dep you’d have to not have a noseto to not notice the stink inating fromEX at the moment but I think at themoment the the AI systems are relativelyU held in high regard however I thinkit’s safe to say that due to theincentives in the space with some sortof great s for AI isinevitable so uhthere are so many things here that areso expensive and so boring that all ofus have completely ignored them butlet’s take a moment to acknowledge thatevery day someone takes a tray out of amachine in a water um system and teststhe water that you drink to make surethat it doesn’t have any toxins in itand that every day uh there arepollution monitoring systems andextensive redundancies in power banksthat make sure that there you never havea blackout that if you’re that if youare ever in a life-threatening situationyou need a power supply it’s there foryou through thick and thin those kindsof reinvestment uh would not necessarilybe provisioned by the private sector umand there’s something that you know justgo invisible and I think that we need tolearn from that um and figure out howcan we take care of the Upstreamcomments just like we look after therivers we need to look after the thelibraries data sets and scientists whoare producing new knowledge no point inhaving an AI based on knowledge if theknowledge dries up just like there’s nopoint in having a water company if theUpstream River dries up similarly weneed to reinvest for Downstream safetyand maintenance costs like continuousauditing and um evals and also just thedeeper work to deeply understand howthese things work um just like we it’snot like we we said oh maybe there’ssome germs in the water and then westopped there The Germ theory of diseaselaunched a research program which iswent still ongoing but if you look atthis uh we all know that machinelearning systems are uh notdeterministic but do we have a thesefirms stopping everything until they’resure that their products are safe no weneed to fund the research for that sortof stuff and not just fund it out of thegoodness of our heart we need to startbuilding up these flywheels that connecttogether Revenue toresilience so that’s my uh the firstpart of my talk here lessons from publicinfrastructure I think these lessonsshow that Public Access infrastructurehas really incredible potential tocreate social value there’s a hugedifference between what the market willdo on its own and what publicinstitutions can do by taking thebuilding blocks of a market firm ofdelivering Goods in exchange for moneyand adding on this extra component ofMaintenance andsustainability that tends the commonsand creat social value