Will A.I. based self-driving Disrupt the Automotive Industry? | Stefan Jenzowsky | TEDxHHL

Will A.I. based self-driving Disrupt the Automotive Industry? | Stefan Jenzowsky | TEDxHHL


[Music] hello everyone ladies and gentlemen we are living in the age of digital disruption we have done it to the stamp the telegram the travel agency the conductors on buses and the lift boys and we tend to digitize everything we can digitize what do I mean by digitizing we basically put it on a computer and/or on the Internet we put we base it on processing power and on smart algorithms instead of human labor so basically we also place it on Moore’s law because Moore’s Law you remember was that old observation it’s by no means the law of nature it’s an observation that computer chips are basically 2 dimensional structures so whenever we increase density on them we actually increase the computing power and an exponential nature and that’s what has been happening with everything we have touched in the digitalization and thereby usually when we digitize something it gets very very fast and very strong so basically when we look at linear growth over time then your growth of performance of anything and we start to digitize it and we you have seen it during your lifetime we have seen it in photography for instance where Kodak Eastman had one of the first teams that actually did digital photography we find that this exponential curve for a while is not as good you have 10,000 times 10,000 export pixels in the photograph you have 100,000 times 100,000 but then finally the exponential growth basically hits a linear growth curve you know from the established industries and that’s the moment people very often refer to as the iPhone moment because after that there’s this chaos and amazement in the existing industries people say we don’t know what hit us why has this changed so fast we didn’t see it coming and many of the old business models basically get fully disrupted the business model of the 36 exposures which were more expensive underneath the Eiffel Tower obviously and if you look today at photography most of the traditional photography businesses have ceased to exist but other things are happening in this area usually when we digitize something we make it free of charge it gets very ubiquitous so if you may want to show me a show of hands who has bought at least 5000 photographs with you today and show me oh I see a lot of hands in the room of course now for a moment imagine that we would have come here today with 5000 photographs each in the analog world right we would have big binders of photographs what are you doing darling I’m going to this conferencing and I’m as always taking all of my photographs with me why on earth are you taking all these binders well because someone may want to show I want may want to show one of these photographs to someone so basically whenever we digitize something it gets omnipresent we are raising the old business models and replacing them by new ones and that’s what I call digital disruption and usually it gets free to use and we invent new business models think of Instagram or others now artificial intelligence is digitizing the automobile driver it’s a digital disruption of the driver end of the steering wheel and this is not evolutionary this is revolutionary when we in Europe started to build these five phases in the traditional automotive industry of self-driving which are now SAE levels the same year Google came up with an automobile that had no steering wheel so this is not evolutionary this clearly is revolutionary it basically jumps over everything we have thought about in the traditional automotive industry and the IT minded folks at Google which were not automotive engineers built this today here in Germany one point eight million jobs depend on the automotive industry a while which is making up about ten percent of our GDP roughly directly and indirectly now interestingly enough ninety five percent of all money that goes into artificial intelligence driving goes to the United States and China in 2015 NOAA confluences have counted how many disruptive startups in self-driving were funded and in the mobility sector they found that more than 95% of the money went to the United States so we are now living in an area where there are huge bets made and it’s very well funded in China and the United States and there are unprecedented numbers way more orders 62,000 cars from Fiat Chrysler I believe no one ever have odd has ordered 62,000 vehicles why would you not enough they ordered another 20,000 Jaguars on top hooba signed the deal for twenty four thousand vehicles GM which bought crews automation now said they were have eighty billion dollars to be invested in self-driving these are very substantial numbers that go far beyond anything we do here in Europe and this basically shows this is a substantial part of these get to be Robo taxis then this will replace a substantial part of the overall taxi fleet of the United States and they’re already rolling with passengers on board so why do they do this why do they go for these large fleets because AI based driving is based on fleet learning since the inception of the Elex net in 2012 which basically shifted the paradigm of artificial intelligence to deep learning and reinforcement learning as the main characters we know that it’s not that important to basically program AI then to Train AI and for training obviously you do most of the training in a computer simulation but you need real kilometres who will miles on the road to this is why we see these large fleets piling up because they learn from the wheel incidents of the world from the so called edge cases that basically teach that cars how to drive now if you look at Google alone which probably is the world leader they started with 1 million miles and June 2015 2 million miles took him more than a year to get to 2 million miles then they went to 4 million miles in about a year later that was 11 2017 they went to 8 million miles less than a year later they are now well beyond 10 million wheeled world miles these are wheel driven miles as opposed to much much more of course in the simulator now Tesla just announced in a nice confluence with industry analysts that they have installed a so-called shadow mode and that they’re trained the AI and shadow mode which means then when every human driver drives a Tesla the autopilot is on in the background so while you’re driving that steering wheel of the Tesla the autopilot also plots out a trajectory for the vehicle and it gets compared which one of the two trajectories are better or match actually if they’re match so I throw the data away but if they don’t match so if the human driver does something else and the autopilot of a Tesla they upload this to the Tesla data center and start analyzing why you drove differently than the car would have basically this gives them the ideal base to train vehicles now in the wheel world and every Tesla Drive in the world now starts to basically build up this in this enormous amount of mileage that the AI network is based on so if you just look at how many miles were driven in the shadow mode so by human drivers training the AI network of a Tesla as they just announced one number and that is that they went 1.6 billion miles in July of 2018 already and this also follows obviously an exponential curve and the rollout of the model 3 as you may know note just started at that point so opposed to all based systems AI systems learned from the fleet and from human input into that fleet they require much less programming and more learning they’re non-deterministic and since alex net a lot of games have been played among them chests among them go if you think about what Google did was deep mind and all of these games gradually time after time were lost by humans the computer which is trained in a deep learning and reinforcement learn temp learning type fashion usually wins against human even in the most complex games we play the only question now that remains is is human driving again and how much time do we really have for that shift so if you look at New York City in the year 1900 you would find this a lot of horse carriages on the streets on Broadway and obviously there was a crazy person on that picture yeah someone who’s living in the future you know some look at this a carriage was out horses crazy right crazy innovator in 1900 the predominant discussions about traffic certain amount horse poo and the amount of horse poo horses would give to new york streets how much time did we have for that seismic shift in that industry well this is 1910 can you see horses well maybe I’ll show you one this is the only last carriage that’s a really old-school person right riding a horse carriage what did happen in the meantime well most of the companies that had to do with horses horse riding horse growing carriage making horse shoe making went bankrupt most of them did not survive that seismic shift because the skill sets were totally different the steel skill sets of raising a horse or making horseshoes has nothing to do with making tires or wheels for an automobile they still the skill sets needed were now petrochemical they were machinery there were gears all the stuff you needed to make a kite even if you made an axle on a carriage the axle would be torn by horses now you needed to make an axle that actually would have talked to it and drive the car itself so basically no horseshoe makers no carriage makers left and the horse took wises gone too in Europe we had a very similar situation I worked for a company that made smartphones when the iPhone arrived and we showed that device to our manager so he said well look at this this is really interesting because it shifts the whole business model right because it basically digitized all the functional other people can give you a function to their phone they can write a so-called app and actually in the future maybe all of the other companies will make apps but not we as a smartphone maker and they said no no no this is not sustainable this will not change the industry when Google imitated very quickly with the Android models obviously they were the two only companies that are dominating today’s smartphone market around the world so do we have a chance in the traditional industries ai driving systems will be 10 times safer and more capable than will based systems today we use rule-based systems mainly in Europe the European traditional car makers believe that these non-deterministic a ice they may never be safe for public roads because they’re not deterministic you don’t know why they made an accident if there is one because they learn from fleet we believe when AI based driving takes over it will be too late for the European car makers and of course regulators and lawmakers hate non-deterministic systems they have no idea how to license them really and now the question was how can we make a start up for a a based driving in Europe how can we drive today’s cars without official intelligence and the first Poorman you encounter if you start a startup company in the midst of Europe that does artificial intelligence based driving that is that you have no car to drive right these cars are relatively closed so we spent our first year we founded our company about three years ago we spent our first year just hacking cars so we were suddenly professional hackers so we bought vehicles and we were so happy when we could drive hands-free this is from a testing ground somewhere see that we hacked the vehicle and we had this driver with friends you know that made us help that gave us help to understand how the vehicle looks like and we started to drive the car by true AI then when we felt more confident about it we actually equipped a regular folks wagon golf without telling folks running we just used it with cameras around it with a NVIDIA supercomputer in the back and we started a test drive to a trip to draw some more attention we drove from Ingolstadt to Wolfsburg without telling Volkswagen and then involves book we announced we had done this and we actually got some press coverage this was out to build this is when the kindest recalled us and said what are you doing there do you want to work with us it looks like you really want to help us to understand how AI base driving works well today we are licensed to be among others to be a supplier to Volkswagen we have a focus wagon a supplier number and these are very nice folks they help us they work with us so long gone are the days when we had to hack cars and we even worked with some of the most beautiful vehicles in the world this is an AI warm or driving case which we did which helps you to drive the car at first in the simulation environment the very nice beautiful Porsche Cayenne which is driven us to a posh facility and then we take it from simulation into the real world this is a drone filming the wielke on exactly the same location [Music] so we train the artificial intelligence in the computer simulation very much like a computer game environment like a baby it starts to learn to drive not to hit other objects and then of course once it has learned all this we can select translate it to the vehicle and regularly it starts at today’s production vehicle to drive itself by artificial intelligence we can go back to simulation back to reality and it should Nash why do we do this well because this nice Porsche car here had the feel it needed an oil change so it is seeking the place where it needs to go to get an oil change it’ll activate itself and it will go there with nobody at the wheel and place itself in the grass facility just in the right place for an oil change [Music] and this is something the traditional car manufactures can relate to they say this makes a lot of sense for us because we have all these repair places and would be really nice so next time you feel that your car needs to go to a garage you may see it you know some of these other vehicles driving themselves by AI remote control by the facility to the repair place with nobody at the wheel now who of you rode a horse to work today because I believe that human driving manual driving will be very much like horse riding in a while we will do it on nice and beautiful days on Sundays you’ll impress your grandchildren Hey look the stick shifting grater and of course we will not outlaw manual driving we never outlawed the horse riding on public roads it doesn’t it’s crazy it doesn’t have a license plate it doesn’t have a backlight and it’s still allowed who have you own a horse today to come here no one because it is not convenient it disturbs traffic and it’s just a thing of the past so what can we do well defending the status quo is no option when the horse people were reacting to the first steam cars in Britain they basically introduced the so-called red flag act which required a human being to wave a white flag in front of the car and say hey danger automobile approaching restricting the car to walking speed in Britain of course this law didn’t hold for long because people were driving cars already in London eh not in London obviously but in Paris and Berlin so they still hope we can still turn the change of events but the consequence is very clear we have to believe in AI based driving we should not fight this disruption which is coming up we should embrace this disruption because it is clearly coming it’s a function of Moore’s law it’s a function of exponential learning fleet learning and we will probably have less than these ten years you saw between 1900 and 1910 til our roads will totally change but now we should start building the new business models very much like starting to build Instagram and not keep selling photographic film thank you very much for your attention [Applause]


Leave a Reply

Your email address will not be published. Required fields are marked *