By Masooma, Sindi Nexhipi, Celina Prosch, Frederik Radvan
The ongoing process of automatization determines the change of our society, as we experience that physical work of all kinds gets reduced and replaced by machines . The fast developing level of technology bares a huge impact on agriculture and economy in general, as well as it affects almost every field of our living standards, due to machines outcompeting the human work force. This immense shift can have numerous reasons, as machines are for example (usually) stronger and more powerful than humans, as well as they work tirelessly, reliable and cost-effective. This are some of the reasons why especially companies consider deploying them.
One good example for a current automatization process is the design of autonomous driving systems, which are a fast-evolving trend our society has to deal with and will for sure outcompete and replace the “old way” of transport – someday. The huge impact of this driverless cars is already observable in the automotive industry, as for example, the vehicle manufacturer Volvo tests the usage of autonomously driving trucks for the long-distance transport .
Hence, it is very interesting that Uber CEO Travis Kalanick announced in June 2015 that he will buy all self-driving cars build by Tesla within 2020 . The company Uber, which was founded in 2009 with its Headquarters in San Francisco, USA, provides costumers taxi services. Additionally and unlike other companies, it offers people on-demand car service and the chance to see exactly where the driver is, as they come to pick you up. Another fact worth mentioning is that Uber considers itself to be a website, so, in essence, Uber claims that it is a digital conduit that connects riders with high quality transportation.
With applying this new technology, Uber would follow the current trend in which automatization is leading us. For this reason it could be and will be an outstanding opportunity for Uber to expand and what once appeared futuristic now becomes reality: vehicles transport humans in a comfortable and safe, as well as self-driven manner. Uber car with installed autonomous driving system and visible camera on the roof  (retrieved 18.11.2016)
Application of Autonomous Driving Systems
The innovative idea to apply self-driving vehicles to the public transport will for sure affect the company and its surroundings. This decision will be a new field for Uber and bares great potential for the use in transport of human resources in general. Being computer-based, these systems shall represent the most rational behavior in any traffic situation. On the other hand, there may arise certain risks due to artificial intelligence and the loss of emotional decision. Autonomous driving car recognizing its environment 
Starting with strengths of autonomous driving, a primary aspect that comes into mind of any individual with respect to car driving: safety. Even though people tend to prefer arriving at their desired destination as quick as possible, it is even of higher importance to arrive safely. In many cases these two options differ by only a few minutes anyway. Using systems that have a consistency in driving and a capability to deal with any traffic situation with ease, independent of mood and daytime, offer a reliable and safe transport throughout the entire day. Co-driving collection of data allows to monitor the quality of driving according to the set parameters of how to best e.g. accelerate the car or take a curve. Analysis of such data ensures the most rational behavior in a daily driving situation.
Being connected to a GPS system the chosen track of the transport service can be optimized when taking into consideration traffic services, such as warning for traffic jams or street accidents. Fast and direct transmission of these data to the autonomous-driving system rather than through a secondary entity like the radio yields higher efficiency of adapting the proposed route and leaves more time for the car to change its track before getting into a time constraining traffic situation. Similarly, knowing the exact position of the car at any time can increase the availability of the transport at a location of higher demand, which is recognized by the system due to direct data transmission.
The aspect of private drivers can already sound risky to some passenger, but as they are getting proved for their reliability via the recommendations of their passengers and they are most probably intending to keep their contract with Uber, there is a low chance of problems resulting from this. On the other hand, the clients getting on the ride do not have to be system identified before they can call their driver, as the service allows pretty anonymous profiles, the driver does not know whom to expect getting into the vehicle. Hence, humans being vulnerable to weapons and potentially threatened by the “new passenger” pose a high risk for becoming victims of a crime. For this reason, replacing the human by a machine without fear or any other emotions and with less vulnerability may be a solution to prevent crimes committed by random strangers in such a transport vehicle to happen at all. On the contrary, in the worst case it could be possible that the computer-based systems being connected to a network get invaded by malware, which means: affected by a hacker attack. Using this technology for purposes of terrorism is questionable, but may result if the system is not being properly protected from potential malware.
From the perspective of the company having a higher number of employees is always related to higher costs. As Uber is a service that only employs part of the drivers and mainly obtains its revenue from mediating between customers and a driving service including public transport, like taxis and private drivers, the aspect of replacing human drivers by computer-controlled ones works only partially. Anyway, developing and integrating the autonomous driving systems requires R&D (research and development), both money additional material resources, financial investments and human resources for optimization. Consequently, the capital asset would appear to be rather costly in terms of time and costs. Nevertheless, this investment could be beneficial over time due to less employees. From the perspective of dismissing employees or just the need of less employees on the other hand can be seen as a negative aspect of saving money, because people would lose their jobs. Whereas looking from another perspective one could say that the R&D would require the creation of new jobs, which would cancel this argument out.
Thinking of a computer as the ultimate rational opponent or tool of the human it did not simply become what it is today on its own. Like other technologies it took decades for computers to work in a convenient fashion as of now. Years were spent on development of the technology and progress was made over the years encountering bug-fixes and malfunctions. Artificial intelligence as a form of autonomous thinking and rational decision making can be a very powerful tool. But, when keeping in mind the creator of this new form of intelligence, one gets back to the error-prone human. This does not mean a human cannot decide rationally, but as it has arisen from nature: a human commits errors. When now transferring all the knowledge and the way of general thinking and rational decision making from a human or even several ones, the input may first of all not be consistent and secondly, contains potential errors, which are to eliminate as one aims for the most rational entity in form of an autonomous, non-living system. Hence, the emotionally independent and rational decision making of the system depends on the algorithms created by human hand. Over time these erroneous decisions could be fixed by specialists or potentially even by the system itself, upon the implementation of machine-learning algorithms. But in general there exists no such thing as perfect rational decision making. Since in some cases the best decision may come from intuition or emotion the universal best decision cannot be rational, especially when interacting with other human beings. Therefore, autonomous driving systems need to be designed carefully and an interaction with the client has to be established that is perceived convenient for the latter during the ride.
One final aspect that could be opposing the use of autonomous driving systems is the absence of another human being in the vehicle. Depending on the individual this may cause fear or higher alertness during the entire drive, which is then perceived more uneasy and inconvenient. It seems likely that most of the older generation will show this kind of concern, as they are not that familiar with (this) new technology and would therefore try to circumvent the usage, since it appears not trustworthy. For this reason, Uber should be prepared to focus more on the younger society because it is considered to be more open towards innovation and the adjustment to it, although the same possibility of mistrust could also happen here. The worst thing that could happen if the technology gets fully applied, is that it leads to the decrease of customer numbers and hence, be a potential threat for the continuity of the company, due to financial shortage and the like. Especially for this reason, Uber should implement the possibility for the customer to decide, whether to use a self-driving car or not, as well as the customer should be able to decide in any case, to enter the autonomously driving vehicle or not. On the other side, this investment could also create a greater independence of the company from partners or private drivers.
Decision making – Notorious Trolley Problem:
A typical example taken for psychological studies of rational and emotional decision making is the “notorious trolley problem”. Most people would recognize this problem in terms of train tracks with people bound to them. One self is positioned at railway switch in power to decide which way the unstoppable train will take and how many people it will kill, because in any of the two options there will be casualties. From the perspective of a human the decision for which an individual opts mostly, depends on how many people would die or whether one self would be the one to blame for having activated the switch causing even more people to die. Obviously, it is a case of rational deciding or of emotional perception.
Applying this theory to an autonomous driving system that is meant to be most rational, an analogous situation would occur if the car break happens to be broken and one´s self was to decide on switching the lane. Here it can be assumed that both lanes are for the same driving direction. Inside the car there is a set number of people and in front there is a pedestrian cross-walk with or without a normal traffic light. Regardless of the color the traffic light shows the pedestrians may still cross the street. Then there can be different scenarios such as having a barrier on the other lane, meaning that the autonomous car would sacrifice itself and its occupants.
Since this problem is pretty complex and not as easy to grab in theory only, there is test called “Moral Machine” one can take to autonomously-driving car itself and make decisions for the corresponding situation given certain conditions. It also offers to create one´s own case. This can give a feeling of how difficult it must be to design an autonomous driving system in order to react appropriately to traffic situation in emergency situations. The test was released by a collaboration with the MIT and can be found via the link underneath. Moral Machine : Decision Making GAME
Sample problem from 
As indicated in the pictures above, the left case would imply the death of an entire female group crossing the road legally, three of them having higher education and the other two being pregnant. On the other hand on the right side, four men and one woman were to be sacrificed in order to save the women on the cross-walk. Out of these four men two are average men, one is a criminal and one is homeless. Most people would opt for the scenario on the right as the women on the cross-walk are all associated with a special status such as higher education and pregnancy.
In short, the decision taken by the computer would strongly depend on its input variables and these have to be put reasonably, because (in general) there is no life worth more than the one of another. Additionally to this, one has to consider that these input variables have to be evaluated very fast, as most of the dangerous situations require an enormously fast decision, which would be less of a problem for the system due to its huge computational capacity. Again, if it comes to rational driving behavior: the systems can only be as good as their programmers. But it can additionally learn from encountered traffic situations, as well as databases and thus improve the quality of the ride.
What would be of more concern is the judgment of all the different inputs, like the different persons status’, which are only based on its own cameras and databases. Hence, solving complex situations could always be one of the main risk factors. For instance, with extreme weather conditions or simply strong precipitation. Here, the method of recognizing the environment in close and far proximity of the vehicle are significant. The use of only a camera based system could bare the risk of a slightly dirty lens or windshield resulting in bad quality pictures used for analysis of the situation. Additionally, sensor systems like radar should be used and adapted to not given false results for interference with water droplets. Ensuring the exclusion of such risks would be a big step forward to an optimized and safe service.
As mentioned before, a computer-based driver can process more data at a given time compared to a human individual. This could enable better availability of the vehicles at places with higher demand and better planned transport routes. The latter combined with more intelligent gear change and optimized acceleration and more convenient slowing down of the vehicle may decrease the amount of fuel needed for the drive. Therefore, money will not only be saved in terms of employing less individuals, but also in terms of lower fuel use and in the end maybe result in less air pollution due to exact planning of most efficient routes, speed etc.
Application of Tracking Systems
Combined with all the other benefits, Uber also offers its customers with a tracking system in its vehicles that in every aspect completes the purpose of what Uber has to offer to its customers. The company Uber is constantly evolving to improve its services with regards to providing a safe and reliable journey and it is for sure not an absurd thought that this gathered tracking data is of great use for the application of the autonomous driving system.
A case reported in Delhi about a woman being raped in Uber cabs tells that Uber does not use standalone GPS systems, but phone-based GPS instead, which is why it is very easy to turn off the GPS by only turning off the phone. These kind of cases can be avoided if Uber uses a standalone GPS, meaning something like black boxes that are implemented in the car itself. Therefore it would be almost impossible for the driver, as well as for the customer, to turn off the GPS signal. A statement that was made by an Uber official in response to this incident states, “We became aware of the incident this morning. Safety is Uber’s highest priority and we take situations like this very seriously. We are working with the police as they investigate, and will assist them in any way we can to determine what happened.” In another statement, they say, “Upon being notified of this incident, our team immediately provided the local authorities with all relevant details, including driver [name, photo, bank verified address] vehicle [license, registration etc.] and trip details” . These statements show that Uber is certainly concerned about the safety of its consumers. In such cases, installing video cameras in cabs might be a good start to help track such incidences and possibly prevent them from happening. This may, however, have the disadvantage of intervening in customers’ privacy while they are travelling, which may not be appreciated by some customers or not be possible to be legalized in some countries, due to legal constraints. Additionally to this, Uber would create a totally new field of transport of human resources and therefore no rules on how to deal with this sitations exist yet. Which is one of the major problems. For example questions arise what to do with all the stored data once it was used, for how long the data is going to be stored, as well as how anonymous the data actually really is etc. This could lead to distrust, as people fear the misuse of their data to non-related services. This is why Uber has to make sure that its privacy policies detail how the company is dealing with all the data, as well as it has to monitor the execution of these policies in order to prevent the distrust in their service.
Possible look of a Car Black Box 
However, a possible investment Uber could do is to introduce black boxes in its vehicles used for the provided service. This could greatly improve the tracking system in case of any accidents. The location of the accident, as well as further information on how it occurred can be gathered upon incorporating special devices into the boxes, which would give immediate information to Uber about the accident. It can be especially helpful, if the accident occurred in a secluded area. Many lives can be saved by having instant knowledge about the crash, so first aid can be immediately delivered to those surviving the accident. Special service of contacting ambulances could be provided by Uber once the occurrence of an accident comes to knowledge of the officials monitoring the tracking system or, which is not too far in the future, algorithms detecting the need of an ambulance automatically send a request for it. This would for sure elevate the faith of the customers in the optimum journey Uber can provide them with.
Most importantly, data can be gathered from these accidents about e.g. which routes are more prone to accidents, especially during certain seasons or events. In this case it would be possible for the drivers, as well as for the autonomous driving system, to evaluate the gathered data in such a way that they choose different routes or be more careful with the usage of routes that have a high accident risk. Looking from the customer perspective, it is easier to plan the travelling and the different routes that should be taken into consideration for optimization of e.g. time, due to the knowledge of the joint tracking and black box data.
Moreover, Uber is going to introduce a new service in the near future, so-called “family-profiles”. The aim of this service is to help connect family members through knowing where other members are travelling. Given that, most people in this highly technologized world own a smartphone, they can easily download the Uber application and use it to track the taxi journeys of their family members and can be informed about their location and safety in time. This also holds true for relatives, who may be travelling alone and there is a risk of safety for them, maybe due to certain neighbourhoods or events. One could even say that the trip is more secure, as the driver also knows that the drive is being tracked by the relatives of the person travelling with them, so he has an additional motivation to make sure the guest arrives at the desired destination. However, this would not be of any concern, if the cab was driven in an autonomous and rational fashion. However, the tracking of the family member via tracking the driver’s phone would only be possible if the car, which does not own a smartphone, signals a GPS signal via the implemented black box.
Generally, the Big Data information Uber gets from its customers through tracking serves as a backbone for this company. It uses this information to adapt to any new demand that may arise and has a high potential to introduce innovative services suggested by its customers. Although it has been banned in a handful of areas, including Brussels and some parts of India, due to the privacy issues of customers, Uber is still a very convenient solution for people, who do not have their personal transport and for example have to commute to work every day . It is especially helpful in emergency situations, due to its reliability in arriving at the right-time and place. It can further improve its services by using all the above mentioned Big Data collection ways and then implementing them in future (autonomous) cab rides given to all of its customers.
Uber – Usage of Databases
A variety of actions is taking place behind the service that Uber provides to its consumers, such as company related decision making, accepting the right drivers or simply brainstorming in order to come up with revolutionary ideas. However, one major component that makes it possible for everything to work properly is the data collected by the company, which then gets organized and stored in a database. Every company needs a database that is organized in such a way that it can easily be accessed, managed or updated.
Uber is a company that offers taxi services to customers through an app available on their smartphones. Thus, it is needed to store location data that is send every few seconds by both, drivers and riders app. As a result, a lot of real-time data needs to be used in real-time. In addition to this, Uber is a globally spread company. This means that different datacenters are present in different parts of the world, which have to interact in a fast and efficient manner. For this reason, the requirements regarding Uber database are to write raw GPS data at very high volume (million locations per second), service reads by time range 40, as well as globally replicated data and SLA (Service Level Agreement) and for Uber this would mean 10ms p99 for writes and 40mn p99 for reads. SLAs in particular establish customer expectations with regard to the service provider’s performance and quality in a number of ways, such as availability, application response, time etc.
So the issue would then be, what is it that Uber uses in order to fulfill all this requirements? Uber found the solution to that by building their own system. This was not difficult to decide, since obstacles, such as financial aspect or lock-in’s were not a problem, because Uber is a very well financed company. (Lock-in’s are locks, as a read lock or write lock, and used when multiple users need to access a database concurrently. This prevents data from being corrupted or invalidated when multiple users try to read, while others write to the database.) Hence, it has access to the top talents and resources that are needed in order to create, maintain and update these kind of complex systems. Thus, they weld together their own system by fusing together two very capable open resource compartments: Cassandra and Mesos. This way it is easier to be dealt with real time data and enabled the company to fulfill the requirements necessary for providing a convenient customer service.
However, there are still some constraints with cross-data center replication. It is cost prohibitive to replicate all locations. The data centers must be able to operate in disconnected/degraded state, meaning that if one of them is not functioning this should not interfere or be of any consequence for the other data centers. Load shedding must also be able to be done while the degradation of a data center. In particular, load shedding is a problem in data stream systems. The incoming rate of stream data is unpredictable, thus when the incoming rate exceeds the processing rate, it is necessary to drop some data in order to make sure that the system load is below the accepted upper bound. Hence, Uber uses some general strategies in order to overcome these issues. Firstly, since writes are cheap, a lot of effort is put in optimizing the reads of the data. All data is written to data center-local keyspace, to write-downsampled stream to replicated keyspace, as well as to the application-level logic in order to configure downsampling – the process of reducing the sampling rate of a signal usually done to reduce data rate or the size of the data – for both, replicated and non-replicated writes.
Putting aside the way how the database works and helps Uber to provide the best service, it is important to also know the way it was upgraded and what are the future plans on making it even better. As mentioned in the beginning, Cassandra was used because of its high read volume, which was used for tracing and trip point time series data. Currently it is used for rider activity feed, dynamic pricing, reliability, versatility, as well as cross datacenters (replication of data). However, Uber is working on fully adapting Cassandra for everything in the future, like real time trip data, invoicing, mapping and analytics.
With this the way of enabling the autonomous driving systems to excess and process the data stored in the databases is already laid and it would only be matter of time of allowing and integrating the autonomous system algorithms into the system that is required and responsible for the service the company already offers.
Summary and Conclusion
It is obvious that replacing the human by a machine, which does not have emotions, would show less vulnerability to crimes happening on the street, as no human sitting in the car could be threatened anymore. However, especially for this reason it is possible that individuals may show fear or higher alertness during the entire drive, due to the lack of a driver and therefore a lack of “security”. The result is mistrust in the technology and disuse. This could lead to a potential threat for the continuity of the company, due to financial shortage. A solution for this scenario would be the possibility for the customer to decide, whether to use a self-driving car or not. Additionally, driver-less cars do not have to be considered as less safe because for example the fact of emotional and therefore unthoughtful driving drops out.
Of course it is costly and time consuming to encounter bug-fixes and malfunctions, as the autonomous driving is a computer based system and therefore, the used algorithms leading the self-driving system are created by human hand and therefore error-prone. This means one has to bear in mind that, due to this fact, even computers do make mistakes, especially when they encounter problems like the notorious trolley problem or the difficulty of computing environmental data. Though, the rate of mistakes could be rapidly decreased upon implementing machine-learning algorithms. This means that algorithms improve their own algorithms and therefore “learn”, which in the end would and will result in a faster outcome/decision, due to the computers huge computational capacity.
However, the system is potentially vulnerable to terrorism related hacker attacks, although this is a questionable scenario and can be circumvented by the company upon implementing strong firewalls. This, for sure, is also a costly investment but necessary for the safety of the customers, as well as it stands against the argument that over time, the entire investment of applying the system to the Uber cars leads to the decrease of employees. The latter is a quite delicate topic, because once the driverless cabs get accepted and fully used by the society this means a lot of people would lose their jobs, especially if the technology is also used for transport via bus, tubes, trains etc. However, people do not have to fear this technology only eradicates certain job fields, as one should look at it from the other side: developing and integrating the autonomous driving systems requires R&D, which is connected to both money additional material resources, financial investments and human resources for optimization. This and the junction and organization of the necessary databases, analysis of the data gathered from both tracking and black box systems etc. are a vast field of new job position possibilities. Anyhow, this investment could only be beneficial over time. Not only in terms of money, but also in terms of connections, as a lot of companies could work together in order to improve their services, which are using autonomous systems.
Summing this all up, if Uber uses its great opportunity and takes all the different variables of this investment carefully into account and cautiously plans their next steps, it is likely that the company will be one of the first trend-setting companies, which are leading us into a world, where indeed “what once appeared futuristic now becomes reality: vehicles transport humans in a comfortable and safe, as well as self-driven manner.”.
All of the stated sources were checked on 27.11.16