AI Annotation Work Changes the Very Relationship People Have Toward the Meaning of Work

The queue is still at 744 tasks. It has been there for the past three hours or so. I've browsed today's news, answered a few unrelated emails, gone to the gym, and even tried to take a nap as I waited, to no avail, for the ticker to go down to 743 or less. Why not just click the Start button and go through Task #744 myself? Because I tried, realized that it wasn't easy, and decided to take a break by unenrolling from it, in the hope that someone else is more passionately up to the challenge, get it out of the way, so that I can work on something else.

Ever since the advent of the AI chatbot age we live in, there has been an increased demand for people to brute-force output, becoming providers of human-generated data that can hopefully make the AI models more accurate and human-sounding. I am no different, joining in the faceless army of AI annotators making incremental steps to improve our technological overlords. As dystopian as all this sounds, the day-to-day of AI training is surprisingly mundane, boring to a degree that would seem unacceptable for those used to working in organizations.

Every job has its repetitive elements, but while in the "real" world those repetitions are unfortunate byproducts of creativity, in the world of AI annotation, the repetition is the whole purpose. By giving AI models slightly different materials of the same type and category, evaluated by people with different backgrounds and personal preferences, the models learn to pick out subtle differences amid the sea of sameness, becoming more precise in their output. While the models advance, their annotators are drowning in tedium, struggling to stay excited doing endless queues of similar tasks.

The current 744-task queue is at least on the less mind-numbing side of tedium. While not allowed to disclose the exact nature of the work, I can say that the range of difficulties associated with different pieces of content, even from the same categories and topic, at least makes it more like navigating urban streets with different obstacles rather than an open freeway with little changes in scenery. The problem, however, is that the mind does not suddenly approach this AI annotation project differently when the expectation of being bored by repetition has been so mentally entrenched.

So even as the time tracker ticks away to signal that hourly wages are still being accumulated, the mind finds a way to get bored. It stops rising to the challenge when a slight difficulty emerges, no longer energized to up the gears to come up with creative solutions, but defaults to freeloading. The invisible army of fellow AI annotators is supposed to be at their desks, some more urgently in need of maximizing their incomes. Why not just stop the timer and get those individuals to tackle the not-so-straightforward? Bored as they may be, at least I can do them a favor by handing over the tough ones. 

As I click the timer off, I find myself scared. What happened to the person who'd feared being called out for only delivering good enough and thus not being able to get new work? As the motivation for improving at work whittled away, I find myself coming to terms with the fact that AI annotation may not just be the latest of many different part-time side gigs that I have done. It is an entirely new animal in terms of rewiring how people think about what fulfills at work, and the relationship with work that is often not about a larger purpose but short-term financial gain.

To put it more specifically, the existence of that invisible army of annotators and hundreds upon hundreds of tasks in dozens upon dozens of queues makes it exceedingly difficult for each worker to even imagine being more than just a tool for generating AL-training output. Physical workplaces can make employees feel like replaceable cogs in a larger machine too, but at least they can get acquainted with their fellow cogs and sulk in solidarity about their facelessness. But when coworkers, if we can fit that in the world of AI annotation, are actually faceless, there isn't even a place to complain together.

I occasionally see cruises and parties for digital nomads to get together, party, and work. I used to not understand it, thinking it was more productive and lucrative to stay concentrated in the quietness of my room. But now I think I understand it a little better. AI annotation, the archetypical digital nomad work of today, gives freedom but takes away identity. Facelessness means never sympathizing and never sympathized. That's why I feel no guilt about leaving someone else I don't know to pick up Task #744, and that's why people are yearning for those physical meetups even if they earn nothing in the process.