The title of the book might make you think of a number of other texts or songs. For me, it matches the syllabication of OZZY OSBOURNE’s “Waiting for Darkness,” so I say it in my head in that same cadence. I also think of the sign held by the guy on the insert of the first TRAGEDY album, which reads “I’m tired of waiting for nothing.” You might also think of other songs having to do with the difficulty of patience, such as TOM PETTY’s “The Waiting,” or KILLING JOKE’s “The Wait.” Truly, though, the reference we are meant to make is to Beckett's “Waiting for Godot,” as Sarah T. Roberts, author of Behind the Screen: Content Moderation in the Shadows of Social Media, explains in the book’s foreword. (Saskia Brown translated the book into English from the original French; as Roberts explains, the title’s pun lands a little better en Français because robot and Godot rhyme in that language.) In that play, the characters wait and wait for Godot, who never arrives. In the same fashion, Casilli tells us that the promises of artificial intelligence are based on a similarly fruitless eternal wait. There is simply no way to do artificial intelligence without involving humans. In other words, there’s nothing artificial involved in the intelligence at all.
Key to the text is the subtitle, which refers to the subcontracting of the work that is involved in making AI seem genuine or real. Through a variety of examples, Casilli demonstrates how the supposedly automatic processes promised or enacted through AI are merely a series of layers of obfuscation involving human labor. He explains in the introduction how “Human workers are not being replaced by sophisticated and precise artificial intelligence applications, but by other humans, who are hidden from view, underpaid, and facing work instability” (p. 6). So it’s not just the concealment that’s the issue. Merely exposing the man behind the curtain would be a cool trick, to be sure, but Casilli does more than that. By revealing that the people who are doing this work are subjugated as they do the work, he makes a strong case for resisting and abstaining from this technology altogether.
In further elaboration of his argument, Casilli makes the case that it’s not robots or machines helping humans with labor, but just the opposite. Human involvement in apparently automated processes is reduced to maintenance tasks that the machines cannot handle themselves. He calls this process taskification, whereby human skills or abilities are whittled down to such a degree as to be mere tasks (p. 19). It makes me think of my past job as a dishwasher, or, excuse me, a dish machine operator. I could spend a few pages enumerating each of the individual tasks I had to do for that job, from busing to rinsing to arranging, before I used the dish machine. Then after it had done its task, I had to dry and store or rewash the dishes. Each of those tasks has many smaller and smaller tasks embedded inside it. Knowing where to put each dish, utensil, or tool is essential, as is knowing which item is most in need by the kitchen staff during a particular shift. My point is that there were only a few items that I ever had to hand wash instead of letting the machine do that work for me. In terms of taskification, all the things I do between closing and opening the machine are unable to be automated. I’m there just to serve the machine. God help me if it breaks down during the lunch rush, too. So, a device as ordinary as a dish machine can reveal the complexity of an allegedly simple task and can also help us understand that the machines train us to do certain tasks to assist their functioning. The human hands cannot be removed from the process. To the diners enjoying their food, the work in the kitchen is mostly hidden, but it’s certainly not automated.
If that manual labor example doesn’t work for you, then consider how Casilli deftly blends the two meanings of digital when discussing the nature of digital labor. It is work both done by hand (tapping, clicking) and done in service of technology. As he says, the taps or clicks are “the smallest of all tasks, perfect for training AI” (p. 13) and these actions are, of course, done with the fingers. This sentence concludes the Introduction and vaults us forward into the rest of the text with the promise of further cleverness and insight as it relates to the modern workplace.
The heart of the book, for me, was the distinction between three types of digital labor: on-demand digital labor, microwork, and social media labor. Each of these forms of work provide different affordances to employers. In the example of on-demand digital labor, there’s an aspect of delivery or courier work that I had never considered before. Not only are bike messengers doing the work of taking a delivery from point A to point B, they are also generating information that helps to train global positioning systems, target advertisements, and inform dynamic pricing (p. 70). These processes all happen even when the couriers do not have deliveries to make or are not “working.” There’s an argument to be made that because they are supplying data to the company by simply moving through space, they should be compensated for producing these data. My jaw fell slack as I read that idea for the first time. There’s a similar explanation of how the passengers of ride-share services also unwittingly supply data used to inform surge pricing by refreshing the app or moving to other locations in search of a cheaper fare (p. 72). Again, the idea here is that there is human work involved in this digital labor (both senses) that is concealed through the use of apps or other platform interfaces. There’s nothing automated about it at all.
The human factor is even clearer in this process when microwork is involved. So, humans decide what food to order for delivery or what destination to reach via ride-share and they generate data as a result. Someone somewhere has to annotate and interpret those data so they can be of some utility to another human. This is known as microwork, where tasks are fragmented into the smallest possible unit of completion and offered up for correspondingly low wages (p. 78). Think of times when you’ve had to disambiguate an image to prove your humanity when completing a purchase online. Then imagine doing that for hours on end for dollars or cents as a reward. Think of trying to parse a poorly encoded voicemail for transcription. Tasks like these, among myriad others, comprise microwork, without which artificial intelligence would cease to exist. Again, it’s not the mere revealing of how this process unfolds that is important. Casilli points out how microworkers “do not benefit from the protection of standard employment, and their rights and skills are not transferable from one platform to another” (p. 103). So, it’s a poverty trap, not a way to move up the social ladder.
Another trap befalls the social media laborers who offer their work for free in the form of blogs (ahem!), restaurant or product reviews, or videos of a-day-in-the-life or video gaming conquests (p. 123). These forms of “hope labor” are pernicious because they almost guarantee that the producers of the text, video, or audio will stay trapped on a certain platform instead of transcending it by getting hired as a professional or becoming successful enough to charge money for their work (p. 123). In all three of these cases, the labor of humans helps to obscure what is meant to be an automated process, and the labor itself is not subject to the same guardrails against wanton exploitation as traditional jobs.
There’s a still more sinister edge to some of the platformed work Casilli explores. The vectoralist nature of much of this work means that those doing the work must use their own devices, vehicles, phones, and Wi-Fi to do the work itself; yet, they do not share in the profits (p. 195). The companies benefit from outsourcing these tools necessary for the job and the workers are the ones who need to maintain them. It reminds me of how rare it is to see a company-branded pizza delivery car anymore. Even in a small business, the owners don’t want to take on that cost when they can put it on to the worker. That same shift happens in many different directions and more harmful ways in the gig economy, even as it helps to make AI seem inevitable and magical. There’s nothing inevitable or magical about it, though. It’s humans all the way down in a race to the bottom.
YOU MAY ALSO ENJOY THESE REVIEWS:
No comments:
Post a Comment