Alex Karp's (CEO Palantir Technologies) vision on AI in the military domain | REAIM 2023

I'm John Thornhill from the financial times and I'm delighted first to be at this conference and second to be interviewing Dr Alex carp the founder co-founder and chief executor executive of palantir Technologies now Alex I think it's fair to say is a somewhat unusual chief executive of a very unusual company uh they're not many Chief Executives I can think of who have a philosophy PhD in who trained under Jurgen herb abermath and who have gone on to run a software company that was launched to apply data and AI to help track down terrorists after 9 11. one of his friends was quoted in the Ft as saying that Alex is allergic to simplicity and you also quoted in that piece as saying the country that wins the AI battles of the future will set the international order so you're in the right place today to discuss all this I wondered if we could start by you telling us what role AI is playing in the war in Ukraine and to the extent you can what role is palantir playing um well first of all very happy to be here thank you for that kind introduction forwarded to my relatives who will be happy to know I'm so accomplished um uh you know that quote just as a reference I think was from three years ago something like that and you know when we um when I was saying this uh when we got involved in what's now called AI in in parts of the dod unnamed farts um because silicon Valley's view of Technology was please send it to the adversaries and send the carcinogens to America and the Allies uh you know if you're getting really rich why should everyone have a good life too um and uh we uh we took a different path which was um and then and so when they abandoned uh the U.S government's efforts to build AI in the war context um we had a lot of trepidation about getting involved because this was like five six years ago go and AI as it was until a couple years ago was really a bumper sticker for fraudsters have fraudsters and people selling snake oil and when you're in software you have to be very very careful not to not be proximate to that because it all looks like a PowerPoint now I could hand out unnamed companies where I know the software doesn't exist and they would have a PowerPoint largely copied they would figure out what we're doing and print it up would look better the people pushing it to you will look nicer better in the American context they have whiter teeth in this context they have a better swim style or something that's very appealing um and so we got involved um in the what's you know digitalization of warfare a number of years ago at the same time uh it we were already still maintaining uh products that we had built largely uh for the west but particularly useful in in in Europe because you have an environment where data protection and anti-terror work has to go hand in hand and that gave us an architecture for understanding uh how an Enterprise would work with segmented data so you just have a huge typically everyone now everyone's excited about uh AI because it's a not a joke it's clearly transformative and they're you're seeing the results of a open Ai and in the consumer internet context I would I would maintain that the history of Technology from certainly in the last hundred years is things coming from the military going to Consumer and that's also what I think primarily actually has happened in AI where and again we're not our role in in in Ukraine has been widely discussed I won't say that much about it but as a matter of theory um the old way of targeting where you don't use uh you where algorithms aren't used is clearly a failure and if you if you go into battle with uh Old School Technology even if you're spending 65 billion a year and you're highly accomplished warfighters like Russia and you have an adversary that uh you knows how to install and Implement uh digital digitalized targeting and AIU obviously are a massive disadvantage but uh which which the world heard and that's why there's a lot of interest in AI um but what's very interesting is the the way we discuss Ai and public networks consumer internet doesn't actually translate uh to private networks so if you're doing digitalization or AI on Health Care records or in the context of War you have issues of you know well the work fighting context you have well who makes the decision under what condition who takes responsibility who doesn't the after an action report if the AI makes a decision who's responsible for that if before War you have the ethics of War who starts the war under what conditions during the war what targets are taking out what kind of targets can be taking out what is a you know every country has an idea of Safe Harbor I.E targets you can't take out targets and need special authorization targets that need different kinds of action action reports you have different kinds of segmentation inside data sources so without going to any details in the Ukraine or other countries you have very different data sources they're structured in a different way you have security issues if you're working in a at War as you as everyone's room would know you can't Supply your all your data sources your asset sources your human interstatement to everyone on The Warfield even by proxy through an algorithm because that can be deciphered so how do you run algorithms and run targeting packages where these things are Paramount so the adversary doesn't immediately know what you're doing how you're doing it and can replicate it and then in the civilian context you know this this is a heavily regulated environment for example in Europe you can't just run inner jurisdictional AI on data sets and Health Care legal Pharmaceuticals in any of the anti-interjurisdictional AI is for lots of good reasons would not be allowed legally or in my view ethically so how do how do you make the AI or digitalization work in that context that's something we've spent the last five years building and what's super interesting now is it was a joke and when I was giving you this quote I largely just got ridiculed by my academic friends but like I don't think I think it's pretty obvious to people in this room not as obvious maybe to people outside this room but the the the ability to dominate The War battlefield on our ethical terms meaning where we can control who does what and under what circumstances will Define who writes the law and how it's interpreted now um this maybe is less controversial in Europe it is particularly controversial I think uh in academic circles in which I lingered but I view it as a banality in a fact and certainly the fact that AI is not an incrementally better technology like maybe mortars and tanks where one country has a slight Advantage it is completely logarithmic and non-linear and so yeah that's a lot but right so just to drill down in that if I am with my journalistic superpowers we'll summarize that as saying what you're saying is that when we talk about wars we talk about kind of mortars and tanks but in fact it's data processing is now the new superpower that is enabling the ukrainians to more than match the Russians well yeah you see yeah the the the the it it used to be that um something like an analytic package plus yes but and what I'm saying on top of it is that um that the the differential lift by Advanced augmented augmented or uh digitalized or AI is so great that powers that know how to implement this build upon it can defeat powers that have been spending 65 billion dollars a year over decades a more recent quote of yours the power of advanced algorithms is now so great that it it equates to having tactical nuclear weapons against an adversary with only Conventional Weapons now is that really true we were hearing from Professor Lawrence Freeman this morning and that um when the war started everyone thought that the Russian cyber offensive would overwhelm the ukrainians take out all of their civilian infrastructure that has largely failed what they're doing is a very kinetic attack on the ukrainians look there's so much about this war that can't be discussed in public I don't think I don't think that's exactly how this went down but people in the room will know more than that about that than I and uh that so I don't think this is actually a fair assessment um uh I and and and and and were that true then I would then you're assuming that the the capacity the capacity first of all there's offensive cyber and there's offensive targeting I'm in the offensive targeting what we build meta constellation is in the offensive targeting business it's a different business but I would so um what what I think is we are at the very very very early days Days of AI but in the last six months um uh we went from a world where AI was and digitalized targeting and in general like ability to take out targets you could call it a digitalized kill chain whatever you want to call it it was a either something for a highly erudite ethics discussion which I find fascinating or something that someone was pitching you to buy where the results were going to be a steak dinner and this has now shifted to your ability to identify the right technology and implement it will determine what happens on the battlefield and I don't think there's any dispute about that in fact one of the major things we need to do in the west is realize this lesson is completely understood by China and Russia and we still spend a very small amount of our budgets on this kind of technology so if you I obviously vary in favor of the general budget staying even growing inflation going more invitation but if you look at budgets in the west what are we spending that's differentiated from our adversaries in in business you have a minality where you try to invest where people don't compete well against you what what we do very well in the west is figure out distributed systems that are powered by software that will allow you to fight this is a very specialized way of building things this is why our software companies have done very well quite frankly most of them from the west coast of the U.S but this is why Battleship battle battle Readiness that works this way is way but what we need to learn is wait a minute what are we focused on who is making the decisions about acquiring Advanced Ai digitalization and what part of the budget goes have they ever implemented something that works by the way very basic things you're going to get bombarded by various software products has the software ever been used in the battlefield should be the first question software is not a theoretical thing I spent the first of part of my life as an academic there are a lot of theoretically relevant questions the first question in a software context is Hughes used it how have they used it what are the results you can't these things take five six years to build if you're very good at it and you need access to the battlefield like you can't build these things in the lab so one of the advantages quite frankly we've had is we've been on the battlefield why have we had that Advantage it's one of these bug feature things because everyone in the Silicon Valley was bereft of Common Sense they were busy exporting things to adversaries and corroding our system with carcinogens in the context of the consumer internet so we got access to the battlefield so how for the people in the room who are actually trying to buy this software how do they know what software Works given it sort of shrouded in secrecy and no one will really oh anybody in this room I presume who's buying software can get on the phone and talk to the people who are using the software that's not their problem the problem is what happens after you find out when it works is there is anyone at the top who's making these decisions talking to the people who are proxy on the battlefield is anyone what portion of your budget goes to it probably 99 of your budget is going to very valuable things and not this kind of thing so you need to have a wider swath by the way you also need a wider swath because you're going to do new things that aren't going to work um okay um this theme of this conference is responsible AI so I'd like to get on to kind of ethics and values which you have mentioned I mean I guess to massively oversimplify in America particularly after 9 11 there was a theory that the ends justify the means that the National Security not our Theory I know we'll come to that in Europe I think it there's a clearly an argument that individual rights are in some respects more important than National Security so which is the bigger problem big Terror or bigger yeah we can you reconcile our give it's always interesting to tell people that things that they will don't believe probably justify we will never believe but are true a lot of the reason why my company succeeded was in the beginning post 911 everyone was like oh Bill find terrorists we don't care how most of our investors are like fine terrorists we don't care how and one of you make lots of mistakes when you do things and I may have made a lot of mistakes one of the best decisions I ever made was reflecting upon Hegel we can have a dialectic we're going to find terrorists and protect data protection and why is that important and by the way we hadn't even bothered selling our product to Europe which is you may you may not know it but almost every European country and some form or another uses our CT product counterterrorism product and why do they use it because if you just find the terrorists and you abandon civil liberties you lose half the society in my view correctly do you want everyone knowing what you do in your personal life I don't and I think it's our god-given right to have parts of our personal lives that are protected you can't but if you don't find the terrorists and this is where I disagree with a lot of my academic friends you're going to have Goose stepping in the streets and I got yelled at all over the world for saying our product palantir has played an enormous role in Europe in stopping the far right not years all these conspiracy theories not because we're in any way involved in targeting far right but because if you protect data and stop stop terrorism you get mainstream parties if you don't protect terror the terrorism you get far right people if you own so you have to do both now getting to the ethics and and what we found by the way is in doing both you end up with something better because in fact you only need access to the data that you have access to and you can do an incredible amount on that data now you get to the AI context where it is a winner take all thing and you have again the debate well we should just throw out our ethics of course our adversaries have by Western standards they do have ethics they just have ethics that are very different than ours and those ethics do not include uh segmenting data so that you can see you can see how the decision make making was done doing AFF before and after uh action reports so you can decide what can be targeted what deciding what drives the what areas AI can be involved in and what areas they can't be and I would tell you the same thing that sounds academic but has made my company very successful when you actually reach for the higher standard you end up with a much more powerful system and you bring everybody in the society with you whereas if you just say we're going to throw out ethics first of all even if you wanted to that is not implementable in any country in this room it's not implementable by the way in America it's like there's a slight myth that in America it's like we're Cowboys we were like yeah we just remember and it is true we're more ready for action than say some other countries but you have you do have very strong codes of Ethics about what you're allowed to do what can be targeted what are the actions what happens if something went wrong who's responsible you know in every Western Society you have a chain of authority that goes from civilian to military there has to be an ability to map that chain onto digitized targeting how do you do that you need and this is where you need an architecture it takes five six years to build which allows you to segment the data and allows you to to expose Parts first of all AI doesn't work in private networks as well as it does and so for for very simple technical reasons you need to have a certain kind of tagging algorithm on the data so when you're doing that tagging algorithm you can also create an architecture impose an architecture that can't be changed by the user on what kinds of data sets they can have and those data sets can be made transparent as inputs into the algorithm and you need to have that as a separate product that works with the algorithm and every western side is going to have to do and by the way so that's the politically kind of progressive version Let me give you the right wing Battlefield position I don't care about this the reason you care about this if you don't care about civil liberties is because that's the only way to know that you are controlling your human and second data and it's not being exported to adversaries in your own organization so if you want to say as a mercurial a matter of theory your Intel Services want to do a special operation with your special services you want to make sure that that every bit of that data is tagged and the algorithms that are feeding into it are not exposed you will need the same architecture that civil Libertarians want exactly the same and one of the most interesting things about running this business for 20 years is civil Libertarians don't understand that they want the same architecture that spies and special operators want for a very different reason they want to come home alive that's how you come home alive okay I'm once I've opened it up to the audience very soon I have one final question that I want to put in before that which is the role of private corporations in this debate a lot of the discussion today has been the fact that we're now in effect Outsourcing a lot of critical National Security decisions to private companies whether it's kind of starlink in Ukraine whether it's to palantir Technologies in Ukraine and elsewhere whether it's Microsoft and what they were doing to protect um Ukraine against cyber attacks and so on so um how can we ensure that private corporations are acting in the public good and what governance mechanisms are there in this area well this there is there is a strong consensus among people in government that governments should decide and we should provide infrastructure there's just a question of does the company actually believe that or not so obviously almost no company in Silicon Valley actually believes that that's why there's all these lawsuits between the European Union and Silicon Valley and you know but we very much deeply believe that Western institutions to decide and we should Implement so the the implementation infrastructure comes down to who controls what happens here I believe that should be 100 the Democratic elected officials and their representatives in the military and other organizations that is a controversial view in Silicon Valley it is a and that has to be verified does the company believe that or not I 100 believe that our products are built to make sure that you the Mandate of power is controlled by Democratic legitimately democratically elected uh officials and their representatives in Military and Intel um there is another more difficult question which is should we build the software ourself I would just say of course there are lots of things you should build and what you see in Ukraine are lots and lots of technical people building on products including ours and that's something that I also support the idea you're going to build the whole infrastructure yourself that I think is unlikely because the cultures that build these products are were cultures of mad men and women and others and managing that crazy group of people is going to be very very hard in a government context but there are also big divisions within those companies themselves on there I mean Google famously did not participate in Project Maven because its employees were against a palantir yourself I mean your co-founder Peter Thiel is a big funder of uh Donald Trump when he was president you yourself and I called him an ass in public yeah yeah I think you said I respect nothing about the dude and that I was vulgar yeah but uh so how you talk about companies as though they're a single entity but in fact they yeah but you're you're doing something which is very fair you're imposing your that our political beliefs on the decision-making and that is a case but at balance here we have a a line and we tell everybody um if you do not feel comfortable so supporting the legitimate efforts of America and its allies in the context of War don't join palantir and so you know that by the way we are discussing recruiting there are a lot of people who don't want to join pouch here because of that I respect them but then this is the wrong company for you and we have we we also won't work you know we don't we've never worked in China we've never sent our products to Russia other people are like are you working sending your products to Russian no well you know so it's it's it's it's just a different beast and we do have huge political divisions around other issues inside pound chair we are a free discourse kind of place lots of people most people a pound here disagree with me on most issues but he and that's completely fine you just but if you want to be a palantir you're and you should know by the way you don't have to work on the military defense our U.S commercial is is the fastest growing part of pound here it's growing like a a weed um or an oak very but it's but you can but you can't then come and say I didn't know and I do this by the way for example in it you know I a German speaker I tell it's like yeah I respect the post-war German pacifism and band and then I've told this to right-wing people I we had people just want to join powder like I just want to kill terrorists well go somewhere else you know it's like that that's not we I really do believe in this dialectic and the people we recruit and retain uh internalize that so that you are both protecting data and finding terrorists but if you're not okay with either as a matter of ideology you just have the wrong company and there are lots of other companies you know talking about the the software is it uh desirable or even possible to regulate lethal autonomous weapon systems it are these things should be regulated and you have two parts the regulation and the enforcement of the regulation both are going to be tough and that's why there's got to be an immediate and concerted effort to look at how that works and that should be very much part of the funding like you know one of the things that's worked very very well in Europe we always have to sign these contracts we can't talk up in in Europe about he uses our product which is slightly annoying because I'm constantly getting yelled at by people who are protected in our product but but um uh but one of the things I really respect about European countries is you had to show that the product could protect data before you could get paid and this is exactly is what should happen in the autonomous semi-autonomous governance targeting perspective you should have to show the product to work in the counterterrorism contact cellular product and counterterrorism product context in Europe you have to show the product works and what does it mean that it works that you can actually find sophisticated actors that are hidden while protecting data otherwise you cannot get the contract it's not a PowerPoint you have to show it working we had one unnamed country where we powered internal external police in Europe and by the time the tender was done no one else competed because they didn't want to show the product working but this is exactly what has to happen in this context okay questions um please can you keep your questions pithy and don't read out your own doctoral thesis uh who has a question we have a bearded guy here at the front thank you very much my name is John Williams holder from the Dutch Ministry of the Foreign Affairs and how do you see the AI competition between the US and China play out in the next five years um there's different parts of the AI competition people usually focus on AI in general terms and are putting two categories together internal and external defense I don't think we are going to win nor do I want to win on internal surveillance China owns that and I I really hope we don't try to do anything similar to that you have actually the closest thing you have to that is your device we don't need more of that we don't need to put that in the Government monopoly of power and so I assume that none of us want to go down that path and I'm certainly not involved I think that's the wrong path then you have targeting and other areas that involve you know Finding targets all the things we've just discussed I think it's proven that we have an advantage but then our Central disadvantage is what are we going to do with our advantage because right now everyone knows we have an advantage and we are just still operating in a world where most of our efforts are focused on Hardware I'm not against Hardware development and we should have more Hardware development and some very large portion of budget should go to Hardware but what needs to happen in the west now is focus like why aren't 10 of budgets on the things we're actually better on that are winning on the battlefield that is we are so far away from that the I would say at maximally one percent I would be more realistic if you said software products that are useful in the battlefield useful as defined by our adversary not by a PowerPoint written by a consulting company but if but what does China and Russia what what makes them what gives them trepidation it's like one percent of global defense budgets in the west or less and if you defined it as something they're afraid of and has been used actually used not your cousins nephews local product that you know that it's it's under a tenth of a percent that has to change yesterday and then people like in this room have to stay involved you know we we have a you know a lot of customers and very different kinds of Industries leaders have to sit on this this is not the thing you can afford to Outsource to the deputy you don't really like which is like one of the reasons we do so well in Tech in America in the west coast is we make Tech a business problem too many institutions make it the problem of the unlikable person I'm not going to talk to if that's if that's going to be your approach to this your country is going to have a huge problem learned you know it's like that's not what the you know the ukrainians do so it's like this has to be a central and personal Focus of the most senior people all the way down to the bottom backed by some it doesn't have to be a huge part of your budget but it can't be one percent it's got to be like five percent six percent because there's got to be some room for failure and it's like if it's a tenth of a percent you can only actually it's not going to work and there has to be some ruthlessness look I'm very pro-sovereignty our product is built to support your sovereignty that's why we have deployments all over Europe and people are waiting on top and about below our product I believe sovereignty is a crucial issue for every country in this room especially the large countries in Europe and I support that but like there also has to be a realistic assessment of what can I get now and what can I get tomorrow and by the way companies like mine are very happy to teach you how to build this stuff but you got to actually buy stuff that's off the shelf as well that exists now all right I can't see any other questions so oh actually there's one right here in the front good afternoon thank you very much my name is my name is Hans Corsa I have a question on uh the sovereignty part where you talked about and about this technology it looks at this technology the the the company who's in front and has the real experience has a major advantage for everybody who has to uh um wants to follow and in Europe we talk about strategic sovereignty yeah and we see that all the technology of this kind is coming from America what should we do um this is a very important and good question first of all get serious up till now the efforts in Europe have been like I I I obviously now I mentioned names but I met someone I very much respect from one of the largest company countries years ago and was like okay yes it comes from America most of us have deep roofed in Europe have you ever talked to any of the people there why is it so hard to come here and work what why is it that like discourse here between people who are you know like I'm pretty Europe affiliate I left most left live most of my life adult life in Europe I wrote my PhD in German I very pro-frenchant in my own way um I'll leave it at that um it's like in America if I want a meeting with somebody they by and large one of me I mean I'm not saying everybody and I'm offensive to some people but like it's like pushing a closed door here and like also you know if you want people entrepreneurs to come here there has to be a special program in Singapore I can get off a plane and I have permanent residency you got to have something like that in Europe like there has to be a serious effort to partner with people like my company and others where we have a cultural affiliation where we believe in the European values but we have different skills with local things and we have on our for our part have to build platforms that you can build on top of be willing to show people how to do it uh and like this is much more like a learned art and like that learned art exists in mostly on the West Coast by the way it's not all of America it's like certain parts but there's like a performative seriousness now I realize it's hard in Europe because one of the things I admire about Europe is the it's radically egalitarian Tech is not radically egalitarian that's just a fact and you that's a very hard thing for Europe to digest I like that about Europe I view myself as a classic European Progressive meaning I still get to say what I think I still have a personal life and I want poor people to have health care and teeth probably would be you know so that like makes me like far left in America and I think probably right-wing in Holland but um uh but uh but um uh um but it's very hard culturally because this is not an egalitarian practice but if you want your industries to be able to compete effectively for tomorrow and provide all the egalitarian benefits that I think most people in Europe cherish and that have been a light to the world you got to get serious about this and serious means like what is it that who does this who can we learn from how can we who's in our local culture very good at this I think France has done some really significant efforts on this um you see efforts that are successful in Sweden there are little offshoots but it's it's you often have the impression when you talk to political leaders here yeah but I could never say that I could never do that in public because it would look unfair Tech is unfair but if you want a fair Society you're going to need some monocom of a better software culture and Europe is so far behind and like it like everyone knows this in private but we have to somehow do better in public because like the whole European experiment which is economic progressivism is crucial and something by the way that I very much personally support but it'll be very it'll be completely called into question if everyone with a high value revenue is sitting in America and it's not good by the way for our Western Alliance like having an alliance where America is providing all the high value revenue and Europe is providing the dollars that's not an alliance it's going to be as strong as what I think we will need to do well against our adversaries get serious Europe I think we have to finish it there but uh thank you very much Dr Cobb for a great talk thank you

Channel: REAIMSummit

Views: 10,650

Keywords: reaim, responsible use of ai, ai in military domain, artificial intelligence, alex karp, alex karp vision, alex karp interview, alex karp ai, alex karp 9/11

Id: L1PnaB15XBI

Length: 33min 19sec (1999 seconds)

Published: Wed Feb 15 2023