after all this discussion about regulation legislation and what's coming up uh from the policy makers let's get a bit more action next coming up is the question ai in europe is it actually in military in europe is it actually smart enough and after that we will discuss some nice strategies around that but i'm very happy to welcome dr urika franke she is and i have to read this again a policy fellow of the european council on foreign regulations a relation sorry more importantly she has an oxford phd which is pretty impressive and she does a podcast that only i think germans can pronounce it's called zeeheit's heiber so maybe we can write it in the chat if you want to search for it and she was awarded german from german magazine capital 40 under 40.
she will tell us or she will give us hopefully the answers of ai in military in europe is it actually smart enough please welcome dr oligar franke hello thank you so much for having me uh today and for this very kind introduction and yeah thank you very much for giving me the opportunity to talk about the ways that ai is used in the military realm it's actually really great that at this ai summit we got a little bit of time to talk about military ai because often the use of artificial intelligence in defense technology or as some may indeed put it the kind of weaponization of ai is a topic that many people who work in and on artificial intelligence aren't necessarily comfortable with and don't want to talk about and so in some circles there's a certain reticence or indeed rejection about this topic but as i will show you over the next few minutes uh this topic really does concern us all so it does concern the ai experts and the developers from
other fields and depending on how the development goes it may indeed concern well pretty much everyone who's listening here today because it impacts warfare the likelihood of conflict and potentially even international balance so thank you very much for the kind introduction as it was already mentioned i i work at the think tank the european council of foreign relations and i work on new technologies in warfare and what they mean for european uh defense and the title of my talk today is military in europe not smart enough and the topic of the talk is the use of artificial intelligence in the military realm why is it interesting to think about this even as non-experts and what should everyone um know decided to use my time to share four observations with you basically four ideas that i hope will be takeaways from you that you remember even after this this talk is done and then i'll be spending about four minutes on on each of those four points which i think is yeah somewhat digestible over zoom
because we all know that this is a bit tricky to to follow so the four points i want to make and i'll tell them to you now and then go into more details uh the following um the first is there are lots of ways that artificial intelligence can be used in the military realm and we actually already have quite a bit of ai in the military realm the second point i want to make is that civilian ai if there even is such a thing given the kind of dual use nature of artificial intelligence and civilian ai can be militarized and and weaponized and i think that's really important to to talk in this in this forum at the third point is that ai is likely and almost certain really to have quite a fundamental impact on warfare and it also has the potential to have a big impact on global stability and that's definitely worth discussing and then fourth and final and the political realm in europe is somewhat unwilling to discuss military ai publicly
with the exception maybe of france and the netherlands and this is despite the fact that this matters for us and that we are indeed already developing these ai enabled military systems so i think that's worth noting all right well let me start with my first point which was that there are lots of ways to use artificial intelligence in the military realm and we already have ai in the military around artificial intelligence as you are all very well aware is a dual use technology some even call it a multi-use technology so this means that this is a technology that can be used in all kinds of different ways both in the civilian and the military valve and like in the like it is the case in the civilian realm in the military realm there are also lots of different uses so ai isn't isn't like one system one technology that that is now being introduced but but more of an enabler um a bit like electricity or um the internet the computer so we're gonna have
a range of ways that ai is is used in the military realm so what are we seeing um one of the the big areas of interest is with regard to intelligence surveillance and reconnaissance that's also abbreviated as isr and here you have basically ai-enabled systems that go through the lots of data that military operations produce and one prime example of this is for instance the hundreds of thousands of hours if not millions that military drones are um producing the kind of video hours that military drones are producing and you don't always want to have a person watch these and then look at these it's going to take ages using ai to kind of shift through this this big data is is extremely interesting from a military point of view another area of application of ai in the military realm is in logistics now logistics is one of those areas that tend to be under undervalued when it comes to military operations but it's actually super important so ai can help here with predictive maintenance so
it tells for instance an airplane when which part needs to be replaced um it tells the military which parts need to be need to be in storage these kind of things efficient shipping autonomous transportation system all very important and it makes military operations cheaper more efficient it's generally really useful ai also plays a role in uh cyber operations um pretty much everything that's kind of possible in the in the real world with ai is is possible and even faster in the cyber realm where there are pretty much no physical restrictions and so you can imagine ai enabled offensive ais viruses and the like but also but also defense command and control another area where i can help to combine intelligence bring it into into one point of contact help allies work with each other and that's something that's really relevant for joint operations swarming is another area where ai can can help to really develop a new military capability so and you may have seen the videos of drone swarm so lots of systems lots
of units for example drones that work with each other look like yeah bird swarms interact with each other and and this may eventually really translate into a new military capability where the armed forces can can surveil a territory with a withdrawal throne for example uh another really important area where ai can influence military systems and military operations quite a bit is when it comes to autonomous systems or semi-autonomous systems so just to give you a sense of this this topic right now about two-thirds of what the united states um is about the united states dod so the defense ministry is investing in artificial intelligence goes into ai enabled systems plus there is also some investment into autonomy uh more broadly so this is definitely an area of focus by many many militaries around the world and then you can also use ai for things such as forecasting where you try to see crises before they they emerge so all of this is actually already happening it's being developed it's being tested i'm going to give
you a few examples of of this as we can see it already at the moment so let's take forecasting right now the the german bulldozer has a project where they look at um at open data information and they try to use ai to help them forecast where crises will develop uh what kind of crises how the bundesweck can react that kind of thing swarms i mentioned that they are actually huge big area of interest the united kingdom for instance has the project that's called many drones make light and that's a work in progress where you have swarms of up to up to 20 small unmanned aerial vehicles so flying drones that operate under the control of just one individual and of course you can imagine here that the opportunity to have one individual rather than 20 could be um of interest to militaries around the world equally the next french fighter jet the rafale so the next version of the french fighter jet raphael will have predictive maintenance and indeed predictive maintenance is already something
that's used in the military aviation industry and this point about the the the civilian um uh aerial industry kind of brings me nicely to the second point i i wanted to make a second takeaway and that is that civilian artificial intelligence can be militarized and weaponized what do i mean by this actually the distinction between civilian and military ai is not as clear-cut as you would think for example the biggest area where swarms that i mentioned earlier are currently being used is actually as firework substitutes you may have seen these youtube videos of of drone swarms in china when um joe biden won the one the u.s presidency they put up a drone swarm that that spelled biden in the night sky so so this is you know an area that that's definitely also of interest for the civilian realm predictive maintenance i just mentioned for civilian civilian aircraft um but also things such as facial recognition language recognition translation services so basically the same technology that allows facebook to tag your friends in your
pictures can also allow the military to identify targets and indeed the the uscia the central intelligence agency in the united states they reportedly are investing in over 140 different ai projects that look into data analysis facial recognition that kind of that kind of thing so again very much dual use something that can be used both in the civilian and the military and it is really this well civilian military nexus and that was one of the reasons why i was particularly happy to have been asked to talk here because in my view this fact is something that ai developers need to take into account even ai developers that have nothing to do with with the military in the u.s we've had this discussion to some extent um in 2017.
so you may remember reports about project maven and the controversy around that so project maven was a project where machine learning was used to come through drone footage and identify hostile actors and this project was a cooperation between the u.s pentagon and google and several several other silicon valley firms and when it became public this was happening there were several google employees that walked out and eventually the program came to an end um but we still actually have a lot of cooperation ongoing between silicon valley firms and and the us military and of course this also happens in other countries and more recent research has actually found that most ai analysts in the uk are are happy to work with the military now to don't get me wrong i'm absolutely not saying that it is a bad idea for ai developers to work with the military but what i'm saying is that it is definitely worth considering for every ai developer how their work can be used in the military realm how it can be
repurposed or indeed how it can be weaponized and misused and because this is the the second point here so ai can can't only be used for you know official state armed forces but it can also be used misused illegally weaponized by actors such as non-state actors so the problem here of course is that even though ai is actually pretty high tech um for most you know machine learning algorithms you need you you spend quite a lot of money to develop working models you need really highly trained people you need lots of data so this isn't something that someone can kind of build in their garage but once an ai model is developed it is more or less just software and software can of course be shared can be downloaded um can can be rapidly the fuse can be replicated uh reproduced sorry several times and this makes it makes it really dangerous um what's the risk here well you can imagine non-state actors so terrorist groups really using ai enabled cyber attacks to attack all
kinds of all kinds of systems um yet you know with drivers viruses or denial of service attacks that kind of attack you can also imagine non-state actors using algorithms to produce fake news uh disinformation deep fake videos uh everything around around that that theme so so there's certainly a warrior here and again my appeal really is for ai developers to keep in mind even if they're developing something that they consider you know amazing and really useful how it can be weaponized and how it can be misused all right so we talked about the military use of ai and said that civilian and military eye is not that easily separable and there's a possibility of weaponization and misuse that become to my two um final points and and uh the the the third point here i want to make is that ai is likely to have a quite fundamental impact on warfare and a potentially big impact on global stability so let me explain um we are pretty much certain that given everything i just said
about the multitude of uses of ai in the military around ai will influence warfare so we will see more more drones forms we will see autonomous systems etc now if it just stays at that level that's of interest to people such as myself it's of interest to people in the military to militaries to politicians etc so they need to make sure that armed forces are prepared to counter attacks need to have state-of-the-art militaries etc but this if it stays at this level it's basically just relevant for experts i'm not trying to downplay it like this may decide whether or not you win a war so it's definitely relevant but it may not concern every single person but there is a potential that military ais the way that ai is used in the military realm may impact global stability much more fundamentally now this is a possibility not a certainty but it's definitely worth thinking about what are the points that i want to mention here now there's a lot of literature on that and i
certainly can't do justice in like three minutes but um just to give you a sense so one one concern that um strategists have is that with more ai enabled systems in the military realm especially more ai enabled sensors and the ability to go through lots of data that was just mentioning we could end up in a situation where the nuclear deterrence that we currently have where we have countries with second strike capabilities that are able to deter any kind of attack on them because they can guarantee that they're gonna fight back that we are undermining the system of nuclear and deterrence why well because ai could help to make the oceans transparent this means that you would have ai enabled sensors again ai using um all the data that are getting in to to to um well that they're analyzing all this data and that this may eventually allow to know where submarines are that are carrying nuclear weapons and this basically means that countries won't have the ability anymore to do second strikes or
to react if they are attacked and this would basically change the way that our current security system operates so that's certainly um dangerous other examples of way that ai enabled military system can really have an impact on global stability is through so-called flash wars so here the concern is that if you have more and more autonomous systems they may react to each other and you may have a situation where there's an incident it could be an attack but it could also just be a misunderstanding really you have an incidence where um you something happens an autonomous ai enabled system military system reacts to that for instance launches an attack another autonomous system of an opponent reacts to that and within a very short time frame you end up in a flash war as we call it so awarded no one no one wanted um the term flash war by the way comes from flash crashes which we know from the financial market where you had these these automated autonomous systems that kind of reacted to
each other in this way and crush the stock market within within a few seconds uh arms races are something else that are really really worrisome i'm happy to talk about this in the in the q a if we have the time but before my time runs out let me very quickly come to the first uh sorry the fourth point fourth and final point i wanted to make and this is an observation of mine that really has come out of the work i've been doing um and that's europe the political realm in europe is somewhat unwilling to discuss military ai publicly and this is despite the fact that well a this really matters for all of us and also that we are already developing these systems so in the united states and in china and also elsewhere but especially in these two countries military ai is an enormous topic so you have significant investment and in fact in the united states you had a military ai strategy well before you had a national ai strategy the
whole debate on ai was really driven by the national security community rather than by um by business community let's say or industry or something something like that in europe that's very different um again this isn't i'm not saying one way is better than the other but this is an observation so so in europe the situation is is different um in so far as most of the european thinking of ai is really driven from an economic point of view some extent a social point of view and i've actually spent quite a long time looking through all the national ai strategies that eu countries have written and so far we have 21 national ai strategies in europe so kind of policy documents published by eu member states and of these 21 documents only a handful i want to say something like five even mention the use of ai in the military alarm and this is even though many of these strategies really cast a wide net and really discuss all kinds of ai applications but they completely
ignore the defense realm this is even true for like big military um actors so italy nato's fifth largest defense vendor does not engage with this topic at all germany has one sentence in their national ai strategy on on defense and so there really isn't isn't that much there the one big exception is france france is the only country in europe that has published a military ai strategy to this day they also designate security and defense as one of the four priority ai sectors industrial policy the french ministry of defense wants to spend 100 million euro per year on ai research so there's definitely a focus there but uh yeah to the states it's pretty much the only european country that that focuses publicly on this as much the netherlands are currently kind of following that lead but but this is basically the current situation and this is and i'll end on this this is the case even though we actually have quite a few um big ticket military development uh items military systems that are
being developed in europe that involve ai and and use ai be it for example the franco german spanish future combat air system or fcas a new fighter jet that is supposed to have drone swarms and a combat cloud we have several um unmanned combat drones and development in france the neuron in the uk the the turan is so all kinds of different different systems and yet even though the industry and the military kind of you know thinking about this and moving along with this on on the political level we don't really talk about this this so much or i would indeed say sufficiently i'm going to end here because i've run out of time let me reiterate the the four points because again i think it's hard to follow um these kind of talks over over zoom so what i what i really wanted you to kind of remember from this is that there are lots of ways to use ai in the military realm and we're already seeing lots of them in development civilian
ai maybe the ai that you are kind of developing can be militarized or weaponized and that's something to to keep in mind ai is likely if not certain to have quite a fundamental impact on warfare and may potentially have a very important impact on global stability and therefore is of interest to us all and finally the political realm in europe remains somewhat unwilling to discuss the topic of military ai publicly despite the fact that this matters for us and that we are developing these military systems and i am basically trying to to change this with presentations uh such as this one so um thank you very much for for taking the time to listening to me and maybe we have a minute or so left for questions thank you thanks a lot holy riker um actually we have some time we have more than five minutes but thanks for this highly energetic talk i was very inspired um there's a lot happening wow maybe you can also share some sources where people who are interested
can read about this maybe in a chat or something later we have actually two questions one is about um possibility of eye-driven terrorism or a hacking military ai i guess that's the same for corporates but here the threat would be even more severe do you know anything about that and how states are prepared yeah so that's that's a concern that was basically the point i was trying to make by um talking about weaponization or misuse of ai um so far we haven't seen many examples of that um but we pretty much know through research that this always happens right so something is being developed by states or by armed forces etc and at one point it basically is gets to two non-state actors um this is true um especially for systems that are easily distributable or can easily be diffused um and and and less true for for systems where this is harder so so luckily we haven't seen you know terrorists with nuclear weapons why well because it's actually really hard to get them
it's really hard to you need a lot of knowledge you need a lot of money and all of that only a few states have them the concern that many many people have is that this is not going to be the case with ai of course it is difficult to develop ai systems and you've probably all seen these these estimations or estimates how much money it took for instance to develop um gpk3 the the um the language um program basically so they estimated a few million so terrorists or non-state organizations won't most likely be able to develop systems such as these themselves but once models are developed once systems are trained um as i said they're pretty much just software and can be can be shared so so yes this is a big concern um it's a concern for instance that terrorist organizations could use ai to well uh fly drone swarms for instance as i was mentioning so they already have drones they may be able to use them in swarms um cyber attacks is
definitely a huge concern um i mentioned this information all kinds of things um are we prepared for that i mean only as much as you can be not not really i'd say um as with with every kind of military system the moment it's being developed someone tries to develop a counter system so so this is definitely happening but but no i don't think that at this point we're really prepared for for non-state actors um weaponizing these these systems against us and this is why we need to keep an eye on you know how to how to diffuse these systems how much to publish this is a big discussion in the in the industry as you may well know how much how much of the the research to make uh to make public um so maybe we're going to see a move away a bit from from very openly shared research to to more secretive development but i wouldn't be able to say thanks for that um there's another question about the conflict in azerbaijan and
armenia and i think we can frame it even a bit more broadly probably we would always affiliate those ai or the ai warfare more with well-developed countries um is is that true or do we see it all over the world happening so um i've actually been looking quite a bit at the at the um conflicts um in in nagorno-karabakh between azerbaijan and armenia and what we're seeing there really is a lot of use of drones and autonomous systems but these are systems that you can buy more or less off the shelf so azerbaijan bought these from israel um so basically this is to say that for me this conflict is an example of a drone war but not necessarily a neigh i enabled for um i think we're gonna have a situation where for a while only really well equipped and and richer states will be able to use ai-enabled systems and indeed within nato and within within europe we're actually worried that the us may do this quite a bit while some european states
especially smaller ones may not have the the financial or technical capabilities to do the same and this could cause interoperability uh problems so in a first stage i think we're going to have a situation that only a few countries will be able to use these systems but eventually um yes i do think that quite a bit of of this will diffuse and proliferate again because this is even though it may be high-tech it can be exchanged somewhat more easily than other systems so so azerbaijan will never get an aircraft carrier because it just doesn't have the the means but but getting um an ai enabled program that allows them to to combine their their drones or indeed getting an ai um enabled cyber um cyber drivers or something like that it's going to be much much easier so yes i think eventually we're going to see we're going to see the fusion and then proliferation but it depends a bit um i'm sorry that i'm not being more clear but it depends a bit on
which which program or which system exactly cool and then thank you very much and and for the end your christmas wish uh what what what you wish that would improve for ai in the military maybe um that other people that are listening should do or could do i mean okay it's a super odd christmas wish maybe but honestly i think i would really like the europeans to just talk about this more openly think this more clearly it's fine if they say they don't want to use or develop or whatever let's say autonomous weapon systems but this approach of basically head in the sand if we don't talk about it it won't be relevant i think that's really dangerous and and not very helpful so under my christmas tree i want loads of military ai strategies i guess i guess my son would agree with you thanks so much dr ricky franco for this and thank you for your time thank you