foreign people yeah um I was going to talk about the tech capabilities and intelligence and defense it says here myths Mysteries and miracles um I must say I'm not an AI person myself so I do know limited things about AI I do know a lot about the application of AI and more about hacking of AI not so much of whatever you know much better than I do but one thing I do know very well because my company Monarch which has been mentioned and some other companies I founded are all defense companies Monarch in particular is a mercenary company for deep Tech Military Intelligence in the west so we know quite a bit about the status of deeptech capabilities in most militaries and intelligence units in the west I know I have limited knowledge about Russia China Other Nation States if you're interested in that we can talk about that as well so the disclaimer here is that most of what I know in this field applies to our Western militaries um so first of all
the myths is um that the um usually in the 80s and 90s movies when you're looking at uh who's usually the the bad innovator who innovates something like a dark AI that dominates the world or is super hex or something is usually the Pentagon the defense industry um so that's the the usual culprit and that's still a belief today among lay people that actually the defense industry and and the defense intelligence Community are very big drivers in Innovation and many of these Technologies and but that is not true I'm afraid um at least there's I mean it's a bit of a checkered picture it's very different in the US and the five eyes for example as in in Europe but by and large and they share the same problems and I have written an article for West Point recently where I Illustrated this as the 3p2 problem and that's payment procurement and political procrastination um so the first P for payment it's all sort of sounds a bit like mundane little problems but together they
really form a very solid block that hinders any kind of innovation so it's not as trivial as it sounds and the first P the payment problem is actually quite um quite decisive so the problem is as you will all very much know that public salaries are not anything that can hire top Tech tenants and so the BND for example when they have when they hire hackers at the moment you can see that in the newspaper sometimes they try to hire Hackers from newspapers and then they pay Tifa udit scene which is uh for for a hacker then he earns a monthly what he usually earns daily in the industry and so it's a payment gap of a factor 10 to 20 to what they pay in the industry if you're a good hacker you don't really have to have a an industry diploma or anything or a University diploma but you've been on hacker conferences then the industry already starts paying you know three to five hundred thousand annually if you're a really good hacker
you'll have an annual salary of 1.2 to 1.5 million at the moment and so that's quite a lot and the government pays to 50 to 60 000 a year so that's not really compared terrible um you have a bunch of people who go there anyway of course because they're loyal and Patriots and they think they should do something for their country and maybe learn something interesting or get to hack the kremel or the Pentagon or whatever and but once you get in there you learn that you're mostly getting lectures on compliance on legal stuff on all the things you're not allowed to do and then they're getting bored and it's not just the payment but also in many cases the culture which is wrong around Innovation and hacking at least and also among towards other deep Tech things and so this is a big problem because that means that in public service we only have very very few uh people who are capable and to understand uh deep Tech capabilities what they imply what uh
the the options that they bring and what the different varieties are and that does not just have the implication that they cannot build it themselves but it also has the implication that they cannot really determine research agendas and they're unable to channel their funding into the right directions because they simply don't know what to fund out of the different options and they're having difficulties in getting their procurement right yeah I have been in talks with the generals who had been so desperate at the lack of capability in hacking that they would just be super happy about just one person or two people writing them the specifications for the procurement of the technology they would like to buy not even that is anything they could get in a very large military apparatus so it's a very desperate situation unfortunately for some very desperate generals and apart from that if you're having on that basic level you're not even having the slightest bit of the level of competence that you require in this higher Innovative fields of
course it's also very difficult to derive tactics and strategies and all the other stuff that you usually need as a military um so now you think that the solution could be to just buy those Technologies from your good old defense industry or from startups because there's a lot of talent out there there's a lot of companies who would love to work with the military because they think oh military batches that's uh millions and billions forever so we're going to have a lot of fun um but public procurement in intelligence and defense is a big mess it's really absolutely awful and terrible and I can tell this is my Arch Enemy now for for eight years public procurement um and just to give you an idea of how big of a message is at the moment when you hand in something at the buy nbv that's my Arch Arch Enemy that's the uh of the German bundes version and forces and so if you hand in something there you have the problem that the first of
all you have to know how to get there so this is this is a process which is so complex that not even a German general knows how to get procurement done in the buy nbv so this is really something that's highly mythical already and but if you managed to get it into the bindbv somehow it takes at the moment despite the fact that they're having 6 000 people working there it takes one and a half years until they take a first look at what you are what you want to buy so that's on in it that's usually already three life cycles uh behind of what you actually want to do and then you're having a very messy complicated process where at least 18 to 20 parties have to sign off the procurement and each of the parties think they know know something and want to change something about the specifications and so whatever comes out of that thing is completely different than what the general or what the Armed Forces wanted to have in most
of the cases and we had very funny cases like a frigate for example which was painted with ink that was water soluble so when it went into water all the ink went off because one of the experts thought this is the cheapest ink and we should buy it but that's how procurement works and unfortunately it's not just up in Germany it's also up in all over Europe and very up in the UK and the United States as well the United States managed to get a model to get um by now and it has taken them 18 years to develop it but they managed to get a model which enables them to do faster procurement the fastest they can go is 18 months if it's conventional procurement that goes through Special Operations because they have a special legal setting and within Special Operations they have a private company that is 100 owned again Bihar that is 100 owned by the special forces the Donovan group and the Donovan group can buy stuff from the market when
they buy it from the gray Market or the black market takes two weeks when they buy it legally from startups they haven't worked before 18 months but the new model they have is that they have developed a system of startups they work with and with the startups they have framework contracts and then from the framework contracts they can draw a relatively quickly Technologies and know-how as well and so that is a model that's working but it has taken them 18 20 years to build it and a lot of political will behind it to to build it and we don't even have the slightest sign of a political will and the moment in Germany and in Europe and to change that and so that means um as a startup entrepreneur when you meet a general on a conference and the general tells you oh wow this is awesome technology I want to have this please come by we need to do this then most of the startup entrepreneurs are yay I have a contract with the
German Armed Forces but that only means that you have an eight percent probability for a contract in 45 years if you stay on it and so that's really very difficult for startups to work with these big actors in in public procurement and big defense companies are used to that so they are used to the fact that any contract takes five to ten years to develop until you're there and they have enough mass to to uh to cover that so they can work with that for them it's okay to just feed their Engineers four five six years until they get the contract because then they get enough money to refinance but as a startup that's just completely impossible nobody's feeding a startup without any client for five to ten years because you tell them what just takes that time to to get into contract so for startups that's very difficult and at the moment there's also no good model to cooperate with the defense industry they've tried a bunch of Defense industry has tried a bunch
of things to work with startups buying them keeping them at arm's length a different kind of contractual relations but nothing really worked most of them ended in drama Fiasco um yeah the final element is a political procrastination and this is more or less a result from the complexity as well so what you would need now if you know that procurement isn't working and that the defense and intelligence machines are not working in terms of getting the Innovation on board that they require you would have to go to a politician and tell them you got to do something and so you go to a politician and tell them you have to do something every now and then you find a politician who would actually like to do something because they know it's a desperate situation and they know that actually the directors of intelligence and the generals want to change the situation desperately because they need the stuff to work and but then as a politician once you start diving into that you learn very very
quickly that you understand nothing about it so it's very high technical complexity it's a very high non-technical complexity and there's a high regulatory complexity and surrounding many of the Technologies already there's many different problem layers there's a few facts there are a lot of gaps there's a lot of people very biased and just lying to your face because much of that is industry driven and they have their own biases and blind spots and things they want to tell you and things I don't want to tell you and so for a political uh decision maker who wants to make a decision in a highly debated Tech field that is a Minefield so because if you're making a wrong decision everybody in the Press will celebrate you as the idiot who doesn't understand his job and makes wrong decisions so every decision maker I met in either in the Ministries or in politics themselves who try to dive into these fields very quickly learned that it's much better to duck away again and try to do something
lies like a household or agriculture and going away from these deep Tech Fields so whenever they they try to fund something because in Germany also in the Ministries you're not punished for not doing anything you're only punished for failure for a failure you publish quite heavily because we have this very risk-averse anti-failure culture still and there's nothing we can change about that and in the short term at least not but you're never punished for not doing anything or just just somehow procrastinating and delegating the decision away from you we call this in hacker circles the the ogp tactic that's the other guy's problem tactics so you're trying to make it some somebody else's problem and then go away from it um but this of course leads to a complete diffusion of any decision making and nothing comes out of it and so on all levels that you would require in defense and intelligence in uh the Personnel that you require to actually demand these Technologies to drive Innovation to ask for it the procurement apparatus
to actually buy these things to drive a market to build a market including the defense industry to some extent and the political decision makers nobody really wants to touch this nobody's able to touch it and that is a very uh general in general very much the situation so there's of course going a lot of funding into AI also into cyber security into my field but it's very generic it's mostly going down a decent cyber security it's going down the very common common funnel so much of the money and cyber security goes to fraunova for example which is completely outlandish so if they're thrown over people here but from our point of view of round over is like 10 years behind reality and mostly even driving completely different role than reality so that's very unfounded and in many cases bad political decision making and that also generates its own gravity because once you have made wrong and bad decisions but you spend a lot of money on it then it has to be the right decision
that's also a thing that's very common in Germany so if you've made a wrong decision but there's a lot of money and a lot of attention then you have to somehow weasel your way around arguing that that was actually the right decision and everybody who says that's not the case is just stupid and so you spent more money on wrong and stupid decisions there's also something that happens some other some very stupid paradigms in cyber security have been held alive for decades and decades despite the problem didn't even exist anymore and so the reality today is now we come to a second part of the talk that in the west AI does not come from defense and intelligence or is at least not driven by defense and intelligence as much as you would think so AI mostly comes from consumer electronics and the advertising industry that's a well-known fact and the defense industry mostly buys its Ai and its cyber security from consumer electronics and advertising um so that has a whole different set of
implications uh you just had the colleague here before me talking about hacking of AI but there's a bunch of different things that you should keep in mind when you're just buying an advertising and Ai and put it into nuclear defense so AI has biases yeah commercial AI for example have has blind spots and gaps a defense engineer would have thought of Defense engineering as a very specific discipline with a lot of safety concerns security concerns a lot of engineering methodologies around that and um in addition many of these AIS are being built I can say that because I'm a high functioning artist 180 on the raddl score and many of the AIS that I know of in the United States are being built by a rather autistic people yeah and many of the programs I know for example that Silicon Valley has a secret program where they hire a lot of high functioning artists into the core design of the AIS Boston has a bunch of programs like that and so that also generates different
kinds of biases I can tell you as an artist we don't know what the you humans are doing in many cases and why you're so complicated about certain things so we design things in a very rational structured fashion from what we believe you want but that may also in many cases not just be the case as we find out frequently in relationships um so in very Dynamic situations like a battlefield in AI which is built with certain biases and built with a certain mindset may get confused and a commercial AI may even get more confused from from marketing if you put it into a better field that's a very common fact by the way of AIS which have been put into battle situations and we had one very funny well for us it was kind of funny a funny example where the Pentagon had two AIS from two different groups in the market and put them into a simulated thermonuclear Warfare scenario so the Russians were starting to shoot something in the sky and their
decision had to be made and Analysis and both AIS instantly decided within microsecond that total Annihilation would be the absolute mathematically preferable strategy and then the Pentagon said we're not going to buy this so that that was something that was giving them a very bad impression and but at the same time this is a very a concerning example because the big Powers China U.S and Russia are all considering approaching their nuclear responses onto an AI due to the speed of missiles so this has something to do with the development of ballistics missile like the Satan II missile for example of Russia which can be shot from the Kremlin just takes a few seconds to be here and put Berlin and nuclear disaster so you have to make a very fast decision and if you first need to read the print out do an analysis ask the general the lieutenant colonel Etc the new origin debt so you need a very rapid response in some of these situations so the tendency automatically given by the speed
of ballistics is to automate it and to have an automated decision process underneath that and then later on as a general you get a printout that your nuclear Armada has just been fired but of course that's something um where all these Powers including funnily also China who has been driving this for a long time are now very very cautious because they learned that there may be quite a lot of room for misunderstandings bad interpretations hacking failures of All Sorts so at the moment there's a lot of discussion in diplomatic circles that you should rather try to block the development of AI and defense and certainly in nuclear responses it should be forbidden to do that in addition in defense and intelligence you will be hacked I mean we didn't we do this all the time and especially AI as the colleague before me has said it's very easy to hack in many cases um so I've been writing about this actually already 15 years ago and it's easily possible to attack the language models the
code the training data to infiltrate the supply chain so these are all things that are very easily possible in the Spy community and which is daily bread and butter business and you cannot protect them against this because they have to be open to a certain extent and many AIS even the military AIS which I know which are being developed here are not being developed under high security conditions so they don't have vetted Personnel a bunch of code is going in and out it's not certified or anything so it's most likely already highly infiltrated and this of course is something where you can if you then consider putting your entire National Defense or even nuclear responses onto an AI which can be quite devastating so that's where we are and um this is basically what's the the lay of the land is in defense and intelligence and intelligence is slightly different of course and because in intelligence mostly used for Austin's analysis and and things like that and we in our company we work a lot
with awesome machines for us it doesn't really make such a big difference which awesome machine you use and which kind of AI the big difference we found out in using AI is the analyst to ask the questions so you have to have a good analyst who knows his job and who knows exactly the target region the target structure and everything and then if you ask the right question you basically get the proper answer out of an awesome AI thing and it doesn't really matter if it's the 10 000 solution or the 6 million solution they're only as good as the human or the information which is actually being still delivered by the field agent and not So Much by uh by AI in most of the cases that we have um but in general um at the moment there's a lot of caution regarding AI I mean even apart from the problem that nobody really understands it and knows how to buy it knows how to fund it in a proper way um there's a
lot of caution and some of the big Powers will of course continue to develop it because they think they cannot afford to lose progress and innovation in this field towards Russia China Etc and so they will continue with very high speed and a lot of pressure especially China because they feel that the United States are threatening them and are very dangerous in in this respect but fortunately we are seeing a lot of caution in actually applying Ai and concrete defense scenarios at the moment it will be applied in some smaller tactical scenarios where it's just about recognizing certain drones rapidly enough or stuff like that but the larger strategic scenarios strategic weapons or strategic decisions makings fortunately we believe will not be the case anytime soon and of course in Europe job unfortunately this makes this a very mundane matter in Europe it's a very hostile environment at the moment towards innovation in Ai and defense and intelligence in many cases there are a few cases which will be toyed around with but by and
large this is not a very friendly environment for the tech community at the moment and that would be my tongue one of the most sobering uh speeches we've had today so thank you for that um we do have five minutes um and I have a question that maybe you know it's quite a doom doom Gloom scenario that you've painted but with a glimmer of hope at the end that at least we're not making uh strategically bad decisions with the use of AI in defense but you talked about the US and China and what about the uh the Israeli defense unit and there's often lots of hype and discussion around startup nation and how a lot of the technology has come from that defense unit is that something that they've done right and what um it's very technical so we in the military we make a distinction between tactical and strategic so nuclear for example is a strategic weapon and tactical is like a Hamas missile that flies into Tel Aviv and and for technical purposes
you can use it um then then you just have standard problems of functionality and Effectiveness in comparison with other other mechanisms that you already have in place and in many cases this is AI it doesn't make things much better or makes things worse and this only applies also to static battle situations like missiles flying into Tel Aviv which is considered a static situation as soon as you have a mess of a more Dynamic situation where there's a brigade here tanks there things moving there drones coming in satellites something is hacked then the AI is usually overwhelmed and confused and you can't use it anymore and that's it thank you very much any final questions from the audience before we move to the app yes uh talking about Israel another case currently quite present in all of our minds is Ukraine do you see any deep Tech capabilities being tested there being implemented there for the first time we know that Ukraine is a quite apt Nation when it comes to I.T maybe there is some
some interesting things you see in that space and while they're doing a lot of things with hacking and the entire Western Arms industry is very happy to have a field testing ground at the moment for a bunch of stuff which they brought to the ukrainians which are happy to fire it in the direction of the Russians but most of that is conventional artillery stuff I mean it's Advanced high-tech things but they don't really have much deep Tech in them the only thing that they are trying to do but not very successfully is hacking um in some Fields but they lack human resources at the moment for that fantastic well thank you again Sandra for an incredible presentation thank you right so before we have a mass Exodus I am going to close uh the final session for today um and I actually met Rana last night and it's one of the most interesting and insightful startups that I've had or should we say scale-ups really I've had the pleasure of meeting recently and it's a