Amazon’s “Artificial AI” Service, Touted by Jeff Bezos, Shutting Down

Amazon is set to decommission its 21-year-old crowdsourced work platform, Mechanical Turk, marking a significant shift in its approach to distributed task management. The tech giant announced Tuesday that AWS Mechanical Turk will cease operations effective September 30, 2026.

“We regularly evaluate our programs, tools and services and make adjustments based on those assessments,” a notice on the Mechanical Turk website stated. “Following an assessment, we’ve made the decision to close AWS Mechanical Turk, effective September 30, 2026.”

Launched in 2005, Mechanical Turk, or MTurk, carved a unique niche by connecting a vast global workforce with micro-tasks, known as “Human Intelligence Tasks” (HITs). These tasks, often paying mere cents, involved activities like data labeling, audio or video transcription, and survey completion – work that was straightforward for humans but challenging for early AI systems. Amazon founder Jeff Bezos famously described the service as “artificial artificial intelligence,” highlighting its role in bridging the gap between human cognitive abilities and nascent machine learning capabilities.

The platform’s name was inspired by the 18th-century chess-playing automaton, the Mechanical Turk, which was, in reality, operated by a hidden human master. Amazon’s original vision for MTurk was to expedite the massive data labeling efforts required to build out its expansive e-commerce catalog. As reported by The New York Times, the company sought to harness human power to organize and categorize the immense datasets underpinning its online retail empire.

At its zenith, MTurk reportedly served over 500,000 workers, often referred to as “turkers.” However, industry observers and workers alike have noted a discernible decline in recent years. Krista Pawloski, a data worker and organizer with Turkopticon, an advocacy group for data laborers, observed that the platform had been “in decline.”

This downturn coincides with the exponential advancement of AI models and the subsequent surge in specialized data annotation startups. Companies like Scale AI, Mercor, and Prolific have emerged, focusing on recruiting and managing workforces specifically trained to prepare data for AI model development. While Amazon itself attempted to leverage MTurk for its own AI initiatives, marketing it as a data annotation source for its SageMaker service, the competitive landscape has intensified.

“Workers are available 24 hours a day, 7 days a week,” an Amazon Web Services webpage previously highlighted, emphasizing the potential for rapid turnaround on human review tasks and labeling jobs. Yet, the platform’s ability to retain its workforce appears to have diminished, with many turkers reportedly migrating to competing services offering more attractive compensation or more structured work environments.

Pawloski attributes this shift, in part, to a perceived reduction in Amazon’s investment in improving the MTurk platform as new competitors entered the market. This evolving dynamic has left many long-time MTurk users seeking alternative avenues for their labor.

The closure of MTurk raises questions about the future of distributed human computation and the quality of data generated by such platforms. A 2023 academic study from researchers in Switzerland highlighted a concerning trend: up to 46% of Mechanical Turk workers were found to be utilizing AI models to complete tasks, potentially compromising the integrity of the data. This finding underscores the escalating challenge of distinguishing between genuine human input and AI-generated output in crowdsourced tasks.

The writing on the wall for MTurk became more evident last month when Amazon ceased accepting new customers for the platform. This move was widely interpreted by workers as a prelude to the service’s eventual discontinuation, prompting many to begin the process of migrating to alternative platforms. Companies, particularly in sectors like insurance and travel, that still rely on MTurk for essential data annotation are now actively seeking replacements to ensure continuity of operations.

For many, MTurk offered a pathway to supplemental income and flexible work arrangements. Pawloski, who began turking in 2008 and eventually transitioned to it full-time, emphasizes the platform’s role in providing “meaningful income” and flexible hours, akin to other gig economy jobs but with the distinct advantage of remote execution via a smartphone. The concern now among these workers is palpable as they face the prospect of losing a significant source of their livelihood.