If you need to use a chatbot to do your job, does that make YOU the robot?

In a well known 2015 blog post, Peter Reinhardt argues that Uber workers can be categorized into two groups. One group (the coders and engineers) runs the system and the other group (the drivers) does what the system tells them to do.

Reinhardt describes the drivers as “Below the API” because their work is orchestrated by lines of computer. (API stands for Application Programming Interface, a protocol standard for transferring data between applications.) Unlike the software engineers who orchestrate the system, drivers function as cogs in the machine. Drivers are summoned to complete the app’s objective function, while the coders “Above the API” are the architects of the process.

Reinhardt’s blog post has been living rent-free in my head for a decade. Today, I think about it every time someone tells me about a breakthrough they just had using a chatbot regularly at work. After all, a chatbot is a user interface wrapped around an API. Almost every app on your devices ping APIs all the time, but chatbots have entangled the APIs of generative AI deeply into the daily working lives of millions people at unprecedented speed.

I have never been an Uber driver, but I suspect I would initially appreciate the sense of autonomy to choose my hours of work. Chatbots, likewise, dangle the allure of personal agency in front of us. How nice to bark orders at a machine that cheerfully complies with my every whim. It feels like power to summon the API to do my bidding. Who wouldn’t want a friendly, benign robot to contend with all the boring, dreary parts of work? It all sounds like empowerment.

Unsurprisingly, AI is marketed as a helpful assistant. It is here to augment our capacity and supercharge our productivity. But in this spirit of consensual reliance are echoes of the dependence that an Uber driver might have on their app. The app, or the algorithm and API underneath it, directs and mediates the driver’s obligations. Sure, the choice to install and open the app is discretionary, but eventually the option to close the app becomes less tenable as one’s performance, commitments and economic livelihood are entwined with the API behind it. Perhaps, over time, the app no longer represents liberation as much as servitude.

The grand value proposition offered by OpenAI, Anthropic, Google and the rest is that AI will usher in a new golden age in human productivity. Which amounts to a palatable way of saying we will become “professionally dependent” on their platform. It is in this sense that we might all become like Uber drivers. While fears about being replaced by AI are rampant, an equally plausible scenario is inescapable, ubiquitous dependence on it. Perhaps similar to the way many of us are chained and captive to many other technologies. Like email, smartphones, and other modern tethers.

But let’s suppose that cognitive human labour remains a prime economic commodity. Let’s suppose we remain smarter than the machines in perpetuity. Let’s suppose the jobs of the future amount to giving directions to agent bots that do the menial, sub-human work we, the higher beings, are too good for. Ironically, being “boss of robots” still consigns human labour to middleware in the stack. We are the biological connective tissue between systems.

Whether we’re smarter or stupider than the robots doesn’t much change the possibility that their presence will make our work more robotic.

On one read, nothing here is new. Civilization is the story of human subordination to our own productivity. We are systemically subservient to oil, fertilizers and fibre optics. Progress is an ever accelerating hamster wheel of accumulative dependencies. This is why comparing generative AI to mechanical looms, electricity, calculators, and shipping containers makes for interesting thought experiments. But history is a timeworn lesson that gains in worker output efficiency and productivity do not necessarily come with any inherent, preordained guarantees of directly benefiting workers themselves.

Today, as usual for new technology, capital owns the means of production. But this time labour is eagerly lined up around the block to be the next to get their paws on the newest machine. The new cage on offer from big tech looks so comfortable we are clamouring over each other to get inside. Few workers seem to be asking: If you need to master AI to do your job, who (or what) is the master of your job?

The sheer concentration of capital investment and the scale of data extraction makes the present AI moment a spectacle to behold. This spectacle of speculation is a race, foolish or not, to be the first to unleash the API that humanity will not be able to live without. The API enthroned as the means of all production. The API that we will all be under. The God API.

Such an economic reality leaves very few people “Above the API” anymore. Hence the race to own the newest, ultimate dependency. Hence the zero-sum lies, gambles and delusions.

But I digress. “Intelligence on tap,” as Microsoft brands AI, will probably become a kind of commodity. Markets will probably figure out that a company that has never turned a profit will not actually spend $1.4 trillion on data centres. There will probably be economic fallout.

And the rest of us will still probably need to go to work.

And we will probably be increasingly dependent on, or “under,” those almighty APIs.