Could Amazon Mechanical Turk workers own artificial intelligence technology?

Automation threatens to wipe out millions of jobs – and even those teaching tasks to computers run the risk of eventually putting themselves out of work

With an unprecedented acceleration in technological development, millions of workers across the world are facing uncertain futures, particularly in the retail sector. Could co-operative models help to provide more security? 

A 2013 study by Carl Benedikt Frey and Michael Osborne highlighted that 47% of workers in the USA had jobs at high risk of potential automation over the next 20 years, particularly workers in transport and logistics.

Another 2016 research by Oxford University and Deloitte suggests that 850,000 jobs could be lost in the UK by 2030 due to automation. The study points out that 1.3m administrative jobs across the public sector have the highest chance of being automated, with the same due to happen with 74% of jobs in transportation and storage, 59% of jobs in wholesale and retail trade and 56% in manufacturing.

In 2016 the International Labour Organisation has also looked at the risk posed by automation to five ASEAN countries (Cambodia, Indonesia, the Philippines, Thailand and Vietnam) and found that around 56% of all employment in the ASEAN-5 was at risk of displacement due to technology over the next decade or two.

Also at risk are workers on the Amazon Mechanical Turk, an online marketplace that requires human intelligence. The platform enables businesses to advertise jobs known as Human Intelligence tasks, which individuals can choose to perform in exchange for monetary payment. By doing so the workers on the human cloud are also helping to train computers to become more human-like.

Can co-ops offer an employment alternative to Amazon?

For example, workers could choose photos similar to a certain picture to help social media artificial intelligence engines to improve their prediction of pictures a user will like. As computers are able to perform new tasks, the jobs on the human cloud are also changing. Workers are paid as contractors and must report their income as self-employed income. A 2016 study by the PeW Research Center revealed that about half of the workers on Amazon’s platform made less than USD $5 per hour.

A group of PhD researchers from three different academic institutions proposes a new model where workers earn ownership of trained AI systems, enabling them to secure long-term benefits from the machine replacing their labour. The recent study – Worker-Owned Cooperative Models for Training Artificial Intelligence – is written by Anand Sriraman from Tata Research Development and Design Centre in India, Jonathan Bragg from the University of Washington, USA and Anand Kulkarni from the University of California, Berkeley, USA.

The authors of the research claim that this model could help reduce the upfront costs of model training while increasing longer-term rewards to workers. The paper is based on the results of a survey of workers on Amazon Mechanical Turk conducted by the authors. The survey revealed that workers were willing to give up 25% of their earnings in exchange for an investment in the future performance of a machine learning system.

Related: How governments can help the platform co-op movement

The survey featured 31 respondents, mainly based in the USA. Nearly half of the workers surveyed said they would be willing to give up some proportion of immediate earnings for a 50% chance to earn twice the forgone amount.

“We believe this royalty model could offer a meaningful alternative to the current system of automation eliminating jobs entirely, letting workers maintain a stream of income from trained models while assuming some of the risk and cost involved in developing a machine learning system,” write the PhD researchers.

The researchers point out that the development of AI training data on crowd labour marketplaces is funded by requesters, in exchange for a fixed price paid to workers for producing the data. They argue that a co-operative model would enable them to choose to accept a fraction of that price in exchange for shares o ownership in the resulting trained system.

The proposed model was inspired by other successful co-operative ventures and activist technologies designed to help to address growing inequalities between sellers and buyers as well as users and platform co-ops like Fairmondo, Stocksy and Loconomics.

So one question for the movement now is, with Platform Cooperativism growing in the USA and across the world, is there a role for co-op retailers to play by supporting worker owned platforms for an inclusive digital revolution?

In this article

Join the Conversation