Page 1

SUNY State-wide Colleges ​many thanks for inviting me to this exciting event when a lawyer takes the stage usually the fun part is over I apologize in advance it's gonna be a fun break though because they're going to be two other speakers at the same time the intersection of law and artificial intelligence poses very important questions that are going to shape the future of mankind in the decades to come and I very much hope to engage you in this important debate now we all know that artificial intelligence is rapidly change how we are living and working routine tasks both manual and cognitive will be automated there are forecasts that robots sometimes called embodied artificial intelligence are going to take almost 1/3 of the jobs and traditional professions by 2025 now that is going to be something that is creating tangents with respect to the law because law is made or was made by humans for humans and questions are going to arise as to for example how to deal with robots great strains in particular are going to arise with respect to robots that are smart that benefit from machine learning that do unpredictable stuff shall we and that's a very fundamental issue treat such robots like humans shall we Accord them legal personality at some point in time now it's clear that law is going to shape the future of artificial intelligence what kind of uses of artificial intelligence are going to be permissible the benefits and costs of artificial intelligence what kind of products and technologies will be in use at what prices and I should like to discuss these issues by putting in an example to you the example of an accident that is caused by a self-driving car now what you see here on this next slide is a screenshot from a video that I'm going to play in a minute that shows a Google that drives in Mountainview on the right hand side and wishes to move into the driving lane and the video is shot from a bus that is traveling behind the car and unfortunately the car makes a mistake because the car assumes that the bus is going to let him in basically to the driving lane and the bus driver doesn't so let's see the video okay so the car was crashed and suffered some damage the boss suffered less damage though and it was the first crash in which a Google car reputedly was involved when driving on public roads so to query on what is the potential of self-driving cars and what are the legal issues that are going to arise in contexts such as these now predictions are that seventy-five percent of cars will be self-driving in about 20 to 25 years there are of course many positive developments associated with this at least four or from the perspective of many we are going to have much fewer car accidents and casualties currently we've got about 1.3 million lives that are lost on the streets worldwide every year ninety percent of these fatal accidents are caused by human faults about 9 percent by environmental conditions and much less than 1 percent by technical defects now assuming these cars work well not like the Google car in this particular example we are going to see a dramatic reduction of the number of accidents we are going to see an increased mobility of persons with disabilities the elderly and we are going to see a significant reduction in the opportunity cost of driving most people spend on average more than four years of their lives driving a car now they're not always driving a BMW which of course gives you only sheer driving pleasure they're driving other cars they are driving they're driving other cars and cannot do things that they might want to do when driving a car like reading or talking and doing other productive stuff now we're going to see fewer cars on the streets because of car sharing people will own less cars of course then today that is going to free a parking space is going to reduce travel time especially in the inner cities and fuel emissions so this is all good unfortunately of course there are negative developments as well we're going to see unemployment of certain people that are employed in certain professions the traditional taxi drivers will vanish bus drivers now you may speculate that if the bus driver now little accident spotted that this was a Google car he kind of deliberately moved faster and not that this Google car get in in order to or you know what I mean and we are going to see him this is an interesting topic a potential higher risk of what's called high magnitude accidents because cars are going to be networked in the future and if we call the cyberattack that might not lead to this one isolated event that we've seen in this car accident with respect to Google but we might see a big big big event where thousands of cars crash at the same time of course the art small accident will also occur all bite rarely so then the question is and I'm going to put this question to you now and we're going to do a pause the question is who shall be liable if eight and that's an important assumption if a fully autonomous car that is self-driving causes an accident and here are five possibilities nobody the loss is worth it for us that's what the lawyers would say the car

manufacturer the device producer if the device producer is different from the car manufacturer which can be the case must not be the case in Google it wasn't the case the car owner there won't be drivers in the future anymore people just own cars if they own a car at all or the car well this might sound very fanciful we hold the car liable but it's not so fanciful as you think when you start reflecting on this particular issue now here is a clipping that I took that relates to an accident apparently that took place a year ago so a robot that did some trafficking on the dog where it was arrested by Swiss police and st. Gallen the robot bought a couple of ecstasy pills also bought a fake I'm Gary and Passport so it was arrested and it was freed after a couple of months by the Swiss police right so what do you think who is going to be the person or the institution or the device that's going to be liable what were we going to see here now a majority if I read this correctly believes that the device producer if different from the car manufacturer should be liable and this is a potential solution it's not the solution that I'm going to think is a solution that should be adopted in the long run and I'm going to make the case now for strict liability of the car manufacturer and those people who work for BMW should not pay attention to what I'm saying right now so nobody liable that's not a good solution of course because that means that expected accident costs would not be internalized by the producers if we've got a defective car that causes an accident which put up wrong incentives of course and also we would then refrain from buying these cars if we weren't sure that in case they caused an accident and hurt us or caused some more severe damage there wouldn't be anybody liable so why is strict liability of the car manufacturer a potentially good solution well strict liability means is no fault required here that means it's just a defective product that causes an accident which suffice is to hold somebody liable and I think this is a good solution because the manufacturer is best positioned to control the risk and Baliff balance the benefits and costs of the technology the manufacturer has intimate knowledge of AI and intimate knowledge of what went wrong or might have gone wrong now this is seen you can see this clearly with respect to the Google car accident Google reacted when the accident happened by saying okay we acknowledge that there might have been some contribution on our side here because something wasn't perfect with respect to the programming of our car now if we didn't think strict lively would be a good idea we could consider fault liability the problem with fault based liability is that could hold a manufacturer liable only if fault were involved and that would require the courts to make a judgment with respect to the standard of care and given these issues I think it's exceedingly difficult to come up with that particular judgment plus a fault liability regime doesn't control what's called the activity level of the producer of a potentially defective device the amount of cars produced and sold and this is also something we want like we would like to do though what about liability of the aia device producer and that's apparently the solution that you came up with in a majority would be the correct solution if that institution is different from the car manufacturer you could think that this is a good idea on the other hand the manufacturer is going to control the assembly process the manufacturer is going to control how the system the complete system is being built and bargains at arm's length with the AI producer supplying these devices so I submit that the liability of the AI device producer is not the best solution rather the car manufacturer if we adopted this particular regime I think we would need to adopt a mandatory product liability insurance regime as well because if we had that magnitude of event which suddenly confronted the car manufacturer with huge huge claims it would be very very difficult to then keep that car manufacturer solvent so there needs to be some sort of liability insurance which is mandatory now you might say this is something that you dream up as an academic and this is true to some extent on the other hand look at this this was a move that Volvo made a couple of months ago announcing that it would take responsibility for the actions of its selfdriving cars I think this is a pretty smart move because it's a very strong signal to the market a marketing signal basically saying we have great confidence in our own technology and it's going to put a lot of pressure on the other management sure to follow a suit because if you don't follow suit apparently your technology isn't as good as the technology evolved oh who can send this strong signal now if we move in this direction there are two complications that I should like to deal with to round up my talk namely the activity level of the owners and how we deal with really smart cars and smart machines in the future what is important of course apart from the care that's exercised by the manufacturer is the activity level of the owner and the owner controls the activity level according to his or her personal preferences the business model I mean that's easy to see if you compare for example a taxi company and a natural person that owns a car there are completely different business models and usage levels and it's also clear that the likelihood of accidents depends on the activity level so strict liability is a problem because strict liability that the car manufacturer ignores the activity level of the owner so we might have to think about something that takes that into account as well a couple of potential solutions here one would be Co liability of owners depending on the activity profile so kind of 50/50 split another one which I think is a little bit smarter it would be that the car when it is being sold comes with a mandatory liability insurance and the premium is determined by a couple of factors such as the type of manufacturer the type of car and the profile usage profile of the owner or the user of the car and things like that are

also available and assisted by the way by artificial intelligence look at this this is called black box insurance it's an English company that offers car insurance and the insurance is tied to a particular device that tracks the personal usage profile of the car driver so we get really very smart insurance rates that take the usage profile of the owner user into account finally and that may be the most intrigued Ewing problem is how do we account for a smart machine learning that is to say things that happened with the card that even though people who create the car are very smart didn't foresee at the time so manufactures really cannot foresee every behavior of smart cars and there is an argument that you could think of to the effect that that might be something that should lead us not to impose strict liability on the car manufacturer but rather and now I'm coming back to this example of the ecstasy pill buying robot rather put liability on the shoulders of the car itself that would mean basically treating smart cars like humans according them legal capacity personhood giving them the power to acquire property conclude contracts etc so is this weird is this science fiction I don't think really when I think about BMW it's a stock corporation a corporation is a legal fiction of course there are humans and the actions of humans are attributed to the corporation but as we're going to see in a second humans are also in firms like BMW are substituted by robots more and more smart cars are capable of two Porsche factions they exhibit what can be called moral or legal agency we could ownership tie to a liability insurance regime like the regime that I've just suggested and treating the car as the target of liability opens up all kinds of interesting new sanctions that we don't have right now reaching from the revoking legal capacity to destroying the car for example so all in all I conclude that in the medium run strict liability of car manufactures coupled with a mandatory liability regime is the sensible approach and in the very long run probably liability of the car it's now as a lawyer I owe you a solution quote-unquote to the Google car accident problem based on these principles Google should take liability to what has happened unless it can be shown of course that the driver committed some kind of wrong in what he or she was doing which doesn't seem to be the case now what I would like to do now is repeat the poll and see whether I've convinced at least some of you or whether I've put you off basically by what I said so here is the paul again and i would like you to again type this into your smartphone who shall be liable if a fully autonomous self-driving car causes an accident yeah great so the car manufacturer comes up the car manufacturer comes up with sixty to seventy percent i stopped here because i fear that my lead is going to melt down if you continue that's enough thank you thank you thank you thank you well two final words the policy road ahead I think the interaction of artificial intelligence and the law opposes crucial policy issues in the decades to come that it should have become clear and a couple of intricate new regulatory issues that have huge societal impact and I've touched upon only some of these issues another one which I mentioned I mentioned it again now is the governance of firms firms are going to be managed more and more not by humans but by artificial intelligence devices the funny thing about all this the interaction of artificial intelligence and the law is that of course the law itself is going to be influenced and is influenced by artificial intelligence artificial intelligence will fundamentally change lawmaking and the legal profession that means for example we are going to have expert systems with respect to specific legal issues legal research and drafting contract formation etc so what emerges on the horizon is something like this we might get smart artificial intelligence based tools to tackle the difficult problems regulatory problems with respect to artificial intelligence that are on the horizon so robots are going to regulate themselves it's just a new form of self-regulation it gets really mind-boggling by the way scientific research and writing can now adays also be assisted by artificial intelligence there are software out there that writes scholarly articles unfortunately I did not find any that would have assisted me in the preparation of this little talk it is all too human that's at least what I believe hence the shortcomings thank you very much for your attention you Bible college (evangelical).