Christian von Borries works as a conductor, composer, filmmaker and producer of site-specific installations. In his work, he appropriates existing music and images and re-samples them. Thus, he high-lightens their reception and appropriation as tools and reflections of social, political and economic control. He is a visiting professor for Intermedia Art at the China Academy of Art, Hangzhou. He has just finished his latest movie “AI is the Answer — What the Question”.
If Artificial Intelligence (AI) is truly Communist (as Libertarian Silicon Valley entrepreneur Peter Thiel claims), then this statistic-based technology could actually compound centralized monopoly capitalism and the coming crisis of inequality as described in Deleuze’s conception of the control society. But it could also be seen and heard as a prototype, a new measure of beauty, redistribution of wealth and solidarity — in short, it could become a utopia that frees us from exploitation, nationalism and racism, from our own perception of this world.
Profit versus the common good
Olaf Grawert: You were able to cooperate with the Future Division of Daimler Inc. as part of the production art festival “Drehmomente” in Stuttgart/Germany: Which ideas collided with each other?
Christian von Borries: The automotive group has been expanding its product range for a while: away from producing cars towards becoming a developer and provider of mobility concepts. According to retired long-serving Daimler CEO Dieter Zetsche, this change in thinking is crucial in the race for the claim of sovereignty over mobility that the company wants to win. Whenever a company like Daimler talks about mobility, it automatically speaks about public space and who is in charge. At this point, private and public interests are blurred, and that’s what interests me as an artist: the relationship between social and corporate interests, the common good, and profit-making intent.
When Stuttgart’s Mayor Fritz Kuhn plows back and forth during a Smart City conference — on the one hand, to reduce inner-city individual traffic, on the other hand, jobs in the automotive industry are more important to the region than “any” mobility concepts — then the failure of the state regulatory power becomes clear. Immediately afterwards, the head of Daimler Financial Services, their rental and leasing division, praises that now everything is based on a single source — from the app to the automobile to the service itself. As an illustration, they presented stills by Daimler’s Future Division: renderings of traffic junctions in Stuttgart, flying cars, people walking on green strips and two men pushing a stroller. I question this narrative of an old future in my practice, which is largely built on associations.
Data instead of taxes
Arno Brandlhuber: At the same time, Daimler’s idea is limited to changing the mechanical world. It replaces workers. Products are no longer produced by hand but by robotic arms. Mobility is not really rethought, the transition to another reality seems unthinkable. An artist talk with you at the Garage Museum of Contemporary Art in Moscow was titled “Algorithms of the smart city and the disappearance of the architect”. Let’s stay with the first part of the title: What does this shift from mechanical to digital algorithmic logic mean for the city, and what role does Big Data play?
CvB: Two aspects are especially important here: Who is collecting which data and who evaluates them? One might think that the utilizer is society, represented by the state, as in censuses. In fact, data generated by the Smart City are mostly amassed by private companies. China is the exception. We all leave our mark on the city: When using public transport, shopping at the supermarket and the like, we are followed by surveillance cameras. So far our actions are not directly related. What we buy at the grocery store, or how many cigarettes we smoke, goes unnoticed without a payback card — unlike our virtual behavior that leads to surveillance and personalized advertising — a circumstance that we are largely aware of and that we seem to have accepted.
In the US and China we can observe a clear tendency towards networks and related protocols of real actions. The key question is: who has which interests in the evaluation of this data of everyday and public life? Speaking of the tax authority, one could imagine a form of justice, concerning public health insurance good motives might play a role. But speaking of private corporations, profit interests are inevitable. The example of self-driving cars shows the change of power from the mechanical to the virtual world very well. Besides Chinese developers, Google is the world leader in navigation technology without even producing vehicles. There is a clear separation between software and hardware, whereby the crucial added value lies in the implementation of the operating system, thus with Google and not with the car manufacturer.
OG: And in the same way, the city is being economized by evaluating and analyzing user behavior. For example, one of the most successful investment funds in the US uses the car park monitoring system of the American supermarket chain Walmart as the basis for its investment forecasts in urban areas. Car brands, car sizes, parking duration and frequencies provide information on the economic and purchasing power and the development perspective of an area. They serve as validation of the trustworthiness of a financial product. This location information is one of the largest and least expensive data sets.
AB: But when the cameras were installed in the parking lots, no one has deliberately decided to give data access to third parties. The information is collected and in some ways exchangeable. Only self-learning software has revealed the importance of data and their availability. One has to ask the question of meaning — where does this qualitative transition happen? What happens if these records — whether random, generated by Google or, in the case of China, government-managed — affect the analogue city and town planning?
CvB: This example shows the successful correlation of data. At the same time we can observe that not software developers but data analysts became crucial. Data analysts do not program applications. These are people who are able to use data patterns to establish relationships in the behavior of people and objects. Do I take the car, the bike or do I walk? Do I leave the house anyway? Who am I talking to on the bus? The old city was planned ignoring the residents insofar as planning and thus responsibility were centralized. For example, where homes and roads are built, how many social housing or private real estate projects are available. The city of the future may be based on population records, a diverse input that feeds the way the city of the future might look.
At the same time, our role is shifting from the citizen to the user. We are no longer part of a society, but a “community”, a homogeneous bubble. The example of Waterfront Toronto shows quite clearly that this economization of the environment by private companies is not perceived as a threat. There, Sidewalk Labs, a subsidiary of Alphabet and sister company of Google, is developing a whole neighborhood. We must ask ourselves how Alphabet’s interests, apart from profit, differ from an idealized urban development.
Data as power
OG: Public-Private-Partnerships are, as the example of Waterfront Toronto shows, on the rise. More and more infrastructure and urban development projects are conceived and implemented according to this model. In Germany, a similar tool exists: town planning contracts. What do these alliances of elected representatives with private companies mean for the city? The withdrawal of the state as we know it?
CvB: It would be too easy to say “because it is Google, it’s bad per se”. Google makes life easier — that’s a fact. And now Google is building an entire part of a city. This step from the virtual to the real world is absolutely logical. For me, the physical presence of the large technology companies at the World Economic Forum 2018 in Davos was a crystal clear indication. For the first time, Google, Facebook and Palantir moved into their own buildings in prime inner-city locations. This was a self-confident spatial expression of power and influence: the companies lined up alongside the nation states, with the difference that access to their representative offices was limited. If you wanted to come “home” there you needed an invitation.
It’s interesting to compare this with China. There, the moral and political difference between private enterprise and state does not exist. Although the big Internet companies are managed privately and traded on stock exchanges, the state has direct rights of access. This leads to a state-controlled information and data centralization. The state has made an important contribution to the development of new technologies such as AI through targeted financial support or participation. Since state structures were previously often perceived as arbitrary, unauthorized or corrupt, today there is less fear of a big brother than of hope for a new form of objectification.
AI is communist
AB: Peter Thiel, founder of PayPal and Palantir, says in your film “AI is the Answer — What was the Question”: “crypto is libertarian and AI is communist”. What does he mean by that?
CvB: Of course, he argues from a libertarian perspective and advocates cryptocurrency as the ideal decentralized system. For him, the idea of central intelligence, as in China, is per se authoritarian and therefore communist. This would make an AI-based and controlled Smart City a la Google per se authoritarian, too. At the same time, there are aspects of a new Cold War, not least over resources, because both technologies consume vast amounts of energy. This has a direct impact on the states involved.
But for Thiel it is also about an idea of the physical society. Their advocates speak of decentralized database-based technology, they always refer to Milton Friedman, the economist and confidant of Ronald Reagan, and his liberal-market approaches that seek to remove any state control. The state is degraded and serves only for the protection of private property. Logically, the next step is to withdraw to islands outside national territory. This new business idea is called “Seasteading” — the epitome of a libertarian society, where any variant of social system is possible: it could be a socialist state, it could also be an authoritarian one.
OG: Algorithms are represented by people who pursue their own ideologies, prejudices and agendas. Among other things, James Bridle writes about the influence of developers and analysts in his book “New Dark Age — Technology and the End of the Future”. This raises the question of which power of action coders and codes have.
AB: In this context, Wendy Chun introduces the sociological concept of homophily in order to explain the effects of algorithms such as those of Facebook. People are divided into homogeneous groups because they are easier to address. Is the Smart City homogenous per se?
CvB: Machine learning is based on statistics. Statistics of user data, existing rooms, situations, environments. However, statistics in market logic also mean that the largest heap is and will always get bigger. Minorities are thus marginalized, which poses a threat, as well as the lack of accountability. Although we know on which algorithms AI is based, the decision levels stay unclear. We can not intervene or disagree, which further contributes to the homogenization of groups and society.
The question is how Facebook would think and understand the city. However, the fact that the Smart City is homogenous per se is highly speculative. Perhaps we should look at physical spaces that Facebook has built for itself and its logic: Frank Gehry’s corporate headquarters design is so interesting because the company relies on one of the best-known trademark architects. But in this case, Gehry didn’t deliver his iconic architectural language. This project is completely untypical for Gehry, but one could argue that he understood Facebook — somehow just as Mark Zuckerberg gets dressed — according to the Norm-core principle: architectural indeterminacy. It’s about the lowest common denominator that works globally and can be reproduced — similar to an IKEA lamp or an H&M T-shirt. The same is true for Alibaba’s headquarter in Hangzhou, which is a boring endless sprawl outside of Hangzhou. On the other hand, just ten years ago, China’s state media broadcaster CCTV hired Rem Koolhaas and Ole Scheeren for its iconic building in Beijing. Nicknamed hemorrhoids, it seems already today rather old-school!
A similar degree of Facebook’s and Alibaba’s vagueness can be observed at Waterfront Toronto. Canada represents a business friendly middle course between the regulatory European and the private-capitalist United States: A hybrid test field that works globally while anticipating a high degree of participation. In Canada, a different type of data generation than in the US is possible — voluntary, pro-active and bilateral. This leads to architectures that on the one hand do not disturb anyone and on the other hand are generated for different and rapidly changing uses.
AB: If we assume that the images of architectures in this case serve to generate resonances, which then are used as data in urban planning, architecture becomes an instrument while at the same time losing its social functions.
CvB: Exactly! First, architecture becomes an instrument of statistics and then provides information about user behavior. The role of the architect no longer exists in this scenario, or it is limited to the design of individual talking points in the urban space that are predetermined by algorithms.
User and Provider
OG: In addition to Orit Halpern, you refer in your lectures to Keller Easterling. You agree that there is a marginalization of certain “undesirable” populations, usually the workers. Your film also shows the Louvre Abu Dhabi, where workers are standing and waiting almost invisibly. What do these pictures tell us?
CvB: For years, I’ve been collecting shots of cleaning staff — uselessly cleaning situations, so to speak. These clips reveal something that is mostly hidden: we all use mobility sharing services, but nobody knows who cleans, refuels, or services the cars. If you change your computer chip or repair your smartwatch, again, the repair worker will remain invisible — invisible and underpaid. Based on such structural conditions, our class system and neo-liberal politics behind it can not last much longer. In already uncertain times, in which societies start to be falling apart — here the rich, there the poor — we must think of these new forms and ideas of future living. But the pressing question will be how established companies and start-ups plan to react to the displacement of the middle class from inner-city areas. Google, Amazon or Baidu may not be interested in this and ideally prevent ghettos of any kind. Could an algorithm counteract this current development?
Big Data as public space
AB: Which tendencies to influence the public space are already evident today?
CvB: Until not so long ago we lived the idea that public space belonged to everyone. The intermediate step is marked by commercialized, semi-public spaces such as the Mercedes-Benz-Platz in front of the O2-Arena, now Mercedes-Benz-Arena, in Berlin. The square was opened by Ramona Pop, the Green Party’s economics senator, with the words: “This is a typical Berlin district as we envision it.” In the neighboring streets, police officers confirmed my question, which was clear to me anyway: No access for the officers on the premises, as it is a private space, and safety is privately thought and regulated.
The World Cup in Russia clearly marks which direction this development is going. There, a face recognition software called FindFace was used nationwide, with a recognition rate of 97 percent. Everyone who visited the FIFA World Cup carried a RFID chip in their passports that had to be carried outside the stadiums as well. This may sound like old technology, but it already anticipates a future in which people identified with chips can cross borders, pay in the supermarket without queuing — that is, live without friction and limits. This form of space and its acceptance marks the transition from big data to public space.
Critics of this technology are countered with the security argument, the synchronization with data of well-known hooligans. Similarly, China argues against criticism of the social credit system. There, big data is already part of the reality of life. Whoever behaves inappropriately, may not use the express train. The reasoning is always concerning marginalized groups that should be convicted. These are just some examples to illustrate how data affects public space.
AB: Sidewalk Labs uses a self-developed open source software called “Replika” to simulate and plan entire cities. They provide it to municipalities and city planners and, in return, receive labeled data to verify their algorithmic forecasts based on real-time database systems. Keller Easterling speaks in her book “Extrastatecraft” of forces that connect you with the physical world. Does that mean that European nation states cannot compete with the superior power of global tech companies?
CvB: Exactly, state administrations have no similar spheres of influence or comparable resources. We need to think about treating and controlling technology companies as supra-state structures. So far, China is the exception, because this one-party-state is organized top-down by default. This is not the rule from a political point of view, but this tendency is inscribed in technologies. Now you might ask: What then is the task for architects in China?
Society and its architecture as an algorithm
AB: Sidewalk Labs CEO, Daniel Doctoroff, was Deputy Mayor in New York and responsible for implementing the communications network LinkNYC, which replaced all phone booths in New York with free Wifi. LinkNYC also belongs to the Alphabet group. This drew sharp criticism from the population. How can one still have a dialogue at eye level in the face of this exuberant economic power?
CvB: If you want to understand the complexity of this business and its reach, you would need to consider a second level and ask: What could be correspondences and social tasks of Alphabet’s technologies and offerings? These would be state responsibilities such as public transport, public hospitals, public health insurance. At Waterfront Toronto, these tasks are going to be performed by private companies. But this is not about traditional privatization, it is about the full access to our habitats and the subcutaneous control of our behavior. You have to understand their business model. In Google’s exploitation logic, all city and state functions appear to be financially free — like a search query. The privatized service is just the tool to get data in return. So the question what tasks architects can still fulfill in this scenario may be too brief. Don’t we have to ask where and how society is actually operating? You cannot act independently of the system. Without Alphabet, Amazon and the like, the situation can hardly be changed. We have to use their tools and think about what we can do with them and what our task should be. You are architects, the software is at your disposal, use it and see what comes out of it, and what that means to you! We are certainly embedded, and there is probably no alternative. In the best case, it extends our utopian horizon.
In a nutshell, Waterfront Toronto might ultimately be the architecture that wins the Pritzker Prize because no architect could ever imagine it. Maybe Big Data gives us an idea of society we would never have come up with.
Supplement June 2019:
The Behavior of Machine Learning (ML)
The programming decisions taken by programmers (such as the value of a learning rate parameters, the acquisition of the representation of knowledge and state or certain connections of a convolutional neural network) influence behavior patterns, i.e. the decision patterns of an algorithm. Coders can influence the behavior of ML by exposing the algorithm to specific training stimuli. For example, many image and text classification algorithms are trained by optimizing the accuracy of a particular record manually using data labeled by humans (labeled data in supervised learning). The selection of the data set and the characteristics that it depicts have a significant influence on this.
The focus on function in machine learning helps us to understand why some behavioral mechanisms of algorithms are spreading and continuing, while others are decreasing and disappearing. The function depends crucially on adapting the behavior to the environment data, not the other way around, as Google’s Sidewalk Labs claims. Successful behaviors (improving the multi-functionality of architecture, for example) are copied by developers of other software and hardware, or further developed to spread on ML algorithms themselves. This dynamic is ultimately determined by the success of institutions such as corporations, hospitals, municipalities and universities — Foucault’s enclosure milieus — who program and use AI to homogenize human behavior.