Jeff Bezos once called Mechanical Turk "artificial artificial intelligence" — a joke that aged into an accidental epitaph. Twenty-one years later, the platform that quietly employed half a million humans to pretend to be robots is shutting down, arguably killed by the robots it helped build.
Two Decades of Invisible Labor
Mechanical Turk launched in 2005 as a marketplace where businesses and researchers could post "Human Intelligence Tasks" — small jobs like labeling images or transcribing audio — for workers to complete, often for just a few cents apiece. At its peak, more than 500,000 workers used the platform, and Amazon will officially close it on September 30, having already stopped accepting new customers on July 30.
For years, MTurk was the unglamorous backbone of the machine-learning boom, generating the labeled datasets that trained the very AI systems now capable of doing the labeling themselves. Amazon's own reasoning for launching it — that "there are still many things that human beings can do much more effectively than computers" — quietly stopped being true for most of what MTurk actually did.
The Robots Ate Their Own Trainers
This is a tidy, almost poetic case of a tool automating itself out of existence: crowdsourced human labor trained the AI, and that AI got good enough to replace the crowd. A 2023 study found that nearly half of surveyed MTurk workers were already using AI tools to complete their "human" tasks — meaning the line between human and machine labor had blurred well before Amazon pulled the plug.
The real successors — Scale AI, Mercor, Prolific — aren't reviving MTurk's model so much as replacing it with higher-stakes work: training frontier models rather than tagging cat photos. The demand for human judgment didn't disappear, it just got more specialized and better paid.
Mechanical Turk spent 21 years pretending humans were machines so machines could learn to stop needing humans. Mission, in a grim sort of way, accomplished.
If your business is wondering where AI can responsibly take over repetitive work without losing the judgment calls that still need a human, that's exactly the kind of integration project we help companies scope out — reach out and we'll walk through it.
Source: Fast Company