Pretend to Be a Bot, Rehumanize the Web

»Maybe it’s all like the German diesel engine: one big lie that will never deliver. After all, we’re already at a stage where people have to pretend to be bots because real bots are just not good enough.« –Sebastian Schmieg

In 2016, Donald Trump’s team hired a Singaporean teenager through the online marketplace Fiverr to convert a PowerPoint into a Prezi, which in an absurd turn outsourced Trump’s »Make America Great Again« campaign – didn’t he want to keep jobs in America? It took the 15-year-old, whose digital marketing services were used to attract younger voters, less than a day to create the file. She didn’t even know who Trump was.
Net artist Sebastian Schmieg takes a closer look at digital labor and the arbitrary nature and strangeness of such online freelance service marketplaces. What happens to us as humans? Do we have to act as if we were bots? Is updating your social media profile already digital labor? In the following interview, gain insight into the artist’s topics and diverse formats, which range from interface performances to neural networks.

Clara Herrmann: One of the defining topics in your artistic work is digital labor. In the abstract for the speculative Prezi »I Will Say Whatever You Want In Front of a Pizza,« which you created this year, you mention digital workers as software extensions, turning around the picture of technology being an extension of our bodies. What is the story behind this work and how is it set up?

Sebastian Schmieg: »I Will Say Whatever You Want In Front of a Pizza« is a piece I made using the Prezi presentation software, in which you don’t simply connect slides, but instead you zoom in and out and rotate everything. It’s a story narrated from the perspective of a cloud worker, and it’s almost exclusively based on found screenshots and videos.

The piece takes a closer look at people working online; having to act as if they were bots; preparing training data for machine learning; designing logos in no time for little money; or simply saying whatever you want in front of a pizza. In this rather bleak context, I optimistically speculate about a group of people who manipulate training data or who plant so called easter eggs, hidden messages or secret features, into the things they produce.

One of the starting points for me when thinking about this was the story of a Singaporean teenager who was hired by Trump’s campaign through the online marketplace Fiverr to convert a PowerPoint into a Prezi. Hence my choice of presentation software.

To me, thinking about digital workers as software extensions rather than the other way around is helpful as an analytical approach. Being a software extension means that you are extending a computational system by offering your body, your senses, or your cognition. This has been true for most kind of work for a long time. However, with software creeping into every aspects of our lives, and with algorithmic systems modulating and optimizing flows constantly, being plugged in and then generating data, or being modulated by data analysis, has become a ubiquitous condition.

To me, it’s obviously important to make a distinction between software and people. However, in a planetary-scale computational system that is constantly optimizing itself toward efficiency and profitability, there is no need for that distinction. Everybody has to be as efficient and valuable a node in this network as they can be.

»To me, it’s obviously important to make a distinction between software and people. However, in a planetary-scale computational system that is constantly optimizing itself toward efficiency and profitability, there is no need for that distinction.« –Sebastian Schmieg

Being a software extension also means that it’s very easy to be removed and discarded. One example: online platforms make it easy for everybody to hire and fire people as needed.

CH: When and why did digital labor become your topic as a net artist? Is this something that becomes quite obvious when working within the digital field and with code?

SS: I think it makes sense to first define what digital labor could mean. A broader definition includes mining rare earth minerals, assembling devices, developing software, etc. Riding a bike for e.g. Deliveroo doesn’t look so digital but it’s part of digital labor, too, as digitally mediated labor. Many argue that updating your profile on social media should be included as digital labor as well.

In 2013, with the artist Johannes P. Osterhoff, I obtained a box of shredded hard drives from a Google data center, which we then, as a readymade, called 10 kg From the New Factory. Obviously, the title is a bit provocative: is the data center really the new factory? But I do think that the infrastructure of the digital is factory-like. Gadgets in our pockets that track our every move, feeding into algorithms executed in data centers, connected to drones and balloons beaming down and sucking up signals, and so forth.

To come back to your initial question, besides making art and teaching, I sometimes work as a freelance web developer, and this way of sustaining my practice has in turn influenced my research and my art. It has made me think about the distributed and globalized nature of development and design, and how everything is constantly being optimized.

And it also prompted me to research online marketplaces for freelancers or entrepreneurs that have created an environment in which outsourcing is cheap and easy for everyone.

CH: Within this context you specifically criticize the exploitative gig-economy, in which only online platforms like Uber or Helping can win in the end without being part of the actual labor itself. »Don’t Just Dream, Do – freelance services for the lean entrepreneur« it says on fiverr.com, a site on which people can set up gigs for a global audience. You also focus on this platform in one of your current projects. How do you try to reveal or subvert the mechanisms of those businesses in your art?

SS: To me Fiverr is particularly interesting as it densely illustrates an environment in which everybody is and has to be an entrepreneur. There, people do not present portfolios or apply for jobs. Instead, users create services for which they can be hired at a fixed price – initially, everything was priced at $5. So it comes down to the best or the most extreme idea, the cheapest service, the most aggressive marketing or the best rating.

Fiverr or Uber are called lean platforms. They only provide the framework, but do not employ any of the workers offering those services. As a consequence, workers themselves are responsible for their own tools, their own training, and their own health-care. They are software extensions, serving other entrepreneurs.

This is not simply a reality limited to Fiverr and similar platforms. The entrepreneur is a model that is becoming increasingly mandatory for everybody. Be the best you can, have a great idea, get funded, work hard, make it to the top, sell, and repeat. And on the way to the top, beat and disrupt all of your competitors. I think it’s safe to say that this produces only a few winners and many losers, increases anxieties and undercuts solidarity among people. Following that logic, you yourself are the only one left to blame if things don’t work out.

Here, I can highly recommend Silvio Lorusso’s research into what he calls the entreprecariat.

In my work so far I’ve addressed Fiverr and similar platforms and developments in several ways. It’s part of the aforementioned »I Will Say Whatever You Want In Front of a Pizza,« and I regularly bring it up when teaching. Once, I had students sign up and work on Fiverr, telling them they would be graded according to the amount of money they made. To me, it’s important to understand the line between »them« and »us« isn’t as clear as one would wish.

»The entrepreneur is a model that is becoming increasingly mandatory for everybody. Be the best you can, have a great idea, get funded, work hard, make it to the top, sell, and repeat. And on the way to the top, beat and disrupt all of your competitors.« –Sebastian Schmieg

In a recent duo-show at Panke.Gallery, I showed a livestream of all videos uploaded to Fiverr. The stream consisted of uploads from people advertising their services as well as completed jobs, uploaded by freelancers, delivered to their clients. Apparently, Fiverr doesn’t care much about securing their users’ data. It’s a rather straightforward piece and I think it offers a valuable glimpse at the globally outsourced media-production assembly line, from glossy animations to personalized porn. Now, the piece lives on as an archive.

CH: You also once said that in today’s Internet, with its social networks, everything becomes labor. Should companies like Google and Facebook pay us?

SS: Yes and no.

The debatable point is whether generating data is labor or not. Playbor? Platforms like Google make money and function as well as they do because they can extract data from everything we do and use this data to optimize their services, and especially their advertising. Hence, Google obviously isn’t built by Google alone, in fact Google search is simply based on the structure of the web as it is.

I really like Laurel Ptak’s »Wages for Facebook« which is kind of a remix of Silvia Federici’s powerful »Wages Against Housework« manifesto from the 1970s.

There are of course qualitative differences to which degree you can speak of exploitation. Nonetheless, I like the demand for wages as a critical perspective when thinking about and dealing with Google, Facebook, etc. I wouldn’t want my friends to send me messages only because they are being paid for doing it. But then, isn’t networking on these sites often a professional practice? And aren’t all the platforms in turn profiting from this, and from all other interactions? The demand for a wage does at least make this aspect visible, and it does ask you if you wouldn’t rather quit.

Furthermore, solving captchas for Google’s reCAPTCHA program makes it very clear how we are in fact working for them all the time, digitizing books, deciphering house numbers, and training their neural networks.

CH: The automatization of labor takes on a new quality with digitization and artificial intelligence. AI and neural networks also play a role in your art. What projects have you been working on, and what’s your position in the discussions within this context?

SS: I started to work with neural networks, the algorithms behind so-called artificial intelligence, when I realized that this is how all this data being gathered can be made productive. Feeding so-called »big data« into such an algorithm allows it to detect patterns, and then to learn concepts represented in the data.

At that time, I was both intrigued and amused by the concept of technological singularity. The first thing I made was sort of an AI recreation of inventor, singularitarian and Google’s resident futurist, Ray Kurzweil. What struck me about his belief in AI and the singularity was his hope for bringing his deceased father back to life. He has archived all of his father’s notes and letters, and they are supposed to form the basis for a reanimation based on AI.

Following this approach, I trained a neural network by having it read all of the books written by Kurzweil over and over again. My version of Kurzweil now lives on a server, where it’s thinking and writing constantly. There, he is a Dadaist of sorts.

»I started to work with neural networks, the algorithms behind so-called artificial intelligence, when I realized that this is how all this data being gathered can be made productive.« –Sebastian Schmieg

The real Kurzweil is trying to prolong his life by taking around 100 pills a day. If this doesn’t work out, then he’s at least still alive on my server, and I own him. I even sold copies to other people so that they could grow and train their own Kurzweil.

Obviously this is an act of appropriation. By researching neural networks I came to understand that a lot of the so-called data sets consist of photos that we’ve shared online. And this data then needs to be prepared for training manually. This is either done unconsciously when e.g. solving Google’s reCAPTCHAs or for little money by people working on Amazon’s Mechanical Turk platform.

I see artificial intelligence as a collective endeavor. We are all working on these systems together. At the same time, it’s as a fight between the big platform companies about who will have the most powerful AI, which can then be rented by everyone else. Again, we are the software extensions, training these centralized neural networks.

With a piece like Segmentation.Network, which plays back all so-called segmentations created by crowd workers for Microsoft’s COCO dataset, I tried to gesture toward the manual labor that goes into building AI, and how this is actually a defining part of what such systems do.

During the past 12 months or so, I’ve worked on a series of connected projects and workshops that address the ethical implications of photographic datasets: what is represented and what is not, and how that in turn shapes the gaze and the decisions that we outsource to machines.

»I see artificial intelligence as a collective endeavor. We are all working on these systems together. At the same time, it’s as a fight between the big platform companies about who will have the most powerful AI…« –Sebastian Schmieg

With the Photographers’ Gallery in London, and all the visitors to their website, I compiled a new conceptual dataset. It’s based on all the photos on their website, including works by Cindy Sherman, Weegee, or Simon Fujiwara. Emulating the way in which datasets are created in a commercial or a scientific setting, people could vote for them to represent either the problem, the solution, the past, or the future.

I’m now using this dataset to make new works, like a computer vision system that can, for example, identify the future or the problem. And I’m also having a neural network synthesize images that somehow represent these concepts, based on all the clicks and photos contained in the dataset.

I’m convinced that machines can learn a lot from photographers.

CH: How do you see AI and automation developing in the context of digital labor?

SS: I think it’s actually quite hard to tell how the whole automation narrative will develop eventually. Of course, businesses are under constant pressure to reduce costs to be competitive in a global market. In consequence, they will rely on machines and algorithms whenever this is cheaper, also compared to outsourcing.

Maybe new and challenging jobs will be created? Who knows. Right now, working for machines is probably as boring and as tiring as it can get. Maybe in the future, it will be like playing a computer game. Which actually doesn’t sound so appealing to me, either.

In any case, this is also a chance to reflect on the status of work. Like »Wages for Facebook« or »Wages against Housework,« this is a helpful perspective. Why are we so obsessed with work? What if we didn’t have to work anymore? What if we want to work but we cannot because we’re not needed? What would we want to do instead?

These questions aren’t so new and in general, the whole situation isn’t entirely new. Artificial intelligence has been around for many decades even if it was in hibernation several times, so to speak. Already Norbert Wiener, basically the father of modern cybernetics, contemplated and also feared the consequences that advanced automation could have for workers.

Maybe it’s all like the German diesel engine: one big lie that will never deliver. After all, we’re already at a stage where people have to pretend to be bots because real bots are just not good enough. At some point, car manufacturers were talking about cloud-workers remote-controlling self-driving cars in situations that were too complicated for the algorithms on their own.

CH: What is your approach concerning creativity and the notion of authorship? Can machines and bots be artists?

SS: No, I don’t think machines and bots can be artists. However, machines and bots can produce art that even surprises the artists who create or utilize them.

To me the question of authorship is twofold. Machine learning, for example, can be considered an act of appropriation. Then, it’s also a question about responsibility. How do you know what a system will produce? Ultimately, a machine or a bot just follows a set of rules or a certain trained behavior. Therefore, I would again say bots or machines do not have any real authorship.

»Do we want AI at all? Probably. So, I guess I’d like to see AI based on collaboration, on curiosity instead of surveillance, on fiction, on small data, on weird data, on ambiguity, etc. I think these are some of the approaches that artists could contribute.« –Sebastian Schmieg

The story behind Horse_ebooks is a nice example of a messy mix of human/machine creativity and authorship. I think it becomes really interesting when the boundaries are blurring and the human-machine relationship becomes messy.

In the context of Fiverr and entrepreneurship in general, I’m often thinking about something which could be called »survival creativity«: coming up with whatever idea it takes to survive in a competitive field. Maybe algorithmic systems are able to develop such a form of creativity, doing whatever it takes to be kept in operation, optimizing everything around such an idea.

CH: How could we think or design AI based on art and artists?

SS: Do we want AI at all? Probably. So, I guess I’d like to see AI based on collaboration, on curiosity instead of surveillance, on fiction, on small data, on weird data, on ambiguity, etc. I think these are some of the approaches that artists could contribute.

Instead of being tedious and exploitative, building an AI could be a communal and playful endeavor which as a practice could us help to think, reflect, and communicate differently.

CH: Besides neural networks, your artistic output encompasses websites, interface performances, algorithmic videos, online interventions, print-on-demand books and now also artistic Prezi presentations. Is the creative adaption of digital tools and software elementary to your work as a Net artist?

SS: I’m not interested in creative adaption as a means to turn something problematic into something sexy. For example, going back to your previous question, an AI based on art or artists should not be something that makes Facebook’s AI sexy, normalizing constant surveillance and prettifying inhumane optimization.

That said, as the invention of a new technology is also the invention of a new type of accident – to loosely paraphrase Paul Virilio – it’s also the invention of a new way of subverting power relationships. My piece »Search by Image« was born out of a specific moment and realization: when Google launched their reverse image search, they had little control over it. Photos of dogs were linked to naked people and so on. This enabled new way of looking at their archives.

It’s also what I’m trying to get at with »I Will Say Whatever You Want In Front of a Pizza,« and with my speculation on Easter eggs, there are ways to subvert these systems that govern us even if it’s only for a brief moment or a quick smile. That’s already something. Now, if there are two or three people enjoying that moment with you, then you know you’re not alone in all this.

CH: But the role code plays in your work has changed over the time …

SS: While I used to be more focused on code as a site for analysis, I’ve come to think that as important as code is, what’s going on there is maybe not just very hard to communicate, but it could also be somewhat of a distraction.

What is there that does not become visible on the surface, especially as code always needs to be performed? As we are all programming all the time – liking things, placing orders, instructing Siri and Alexa, and so forth – the question is how are we being programmed, how do we program others, and how do we program ourselves?

»As we are all programming all the time – liking things, placing orders, instructing Siri and Alexa, and so forth – the question is how are we being programmed, how do we program others, and how do we program ourselves?« –Sebastian Schmieg

Most of the times, coding is about efficiency, speed, optimization etc. It’s about getting rid of what’s deemed superfluous. So it’s a question of ethics and of ideology.

I realize this is somewhat of a non-answer that’s even contradicting itself. Maybe I can put it like this: I’m less interested in the inner workings of a machine than I’m interested in the kind of machines that we are turning ourselves into.

CH: Solitude juror for web-based media Nishant Shah wrote that art may have the capacity to rehumanize the web. How would you like to put the human back in the center, also within your art?

SS: Being skeptical toward the propaganda of technological progress is a first step. And we need ways that allow one to say »fuck you.« It’s liberating, and it’s human.