Rightly or wrongly, the AI-driven world has come to typify the woes of modern economic life, as tech giants such as Facebook and Alphabet amass vast fortunes, due in large part to the huge quantities of data that users often freely provide them.

Accusations of economic imbalance tend to be multi-faceted.  Not only do these companies typically employ fewer people than the industrial titans of yore, but they also attract the ire of officials over their tax practices and have grown insanely rich off the back of something users receive no compensation for.

It’s helped to create a world in which the haves are increasingly well off, while the have nots make do with insecure and poorly paid work.  Nowhere is this exchange more evident than in the data annotation industry, where people from around the world help to prepare and tidy up the data used by the tech giants to train the algorithms upon which their fortunes increasingly rest.

The hidden bedrock of AI

These individuals are the hidden champions of AI, providing the invisible work that underpins much of the breakthroughs achieved in machine learning applications around the world today.  These people are very much the poor relations of the tech world however, recruited to projects via platforms such as Mechanical Turk or Appen, and often paid a pittance for their labor.

While the work performed often involves relatively menial tasks, such as identifying the objects in an image, or providing accurate labelling, the importance of the work was perhaps underlined by the $300 million valuation placed on Appen when it was bought by their larger rival Figure Eight in March.

It’s a world that can easily be characterized as a race to the bottom, with platforms seeking every cheaper sources of labor to do work that tech companies scarcely care as to the origins or conditions the work is completed under.

On the surface, the leafy environs of Ashridge House in the UK seems an unlikely place to find a change in this approach.  The historic English country house is the home not only of the Ashridge Business School, but of the Hult Prize Accelerator, which attracts startups from around the world to compete for $1 million funding to take their idea to market.

Ethical data

Among the finalists are Cambridge University based startup Meaningful, who have joined companies such as iMerit, Samasource and CloudFactory in wanting to make AI data work more dignified.  They believe that there is an opportunity to offer incomes more commensurate with the living wage in the UK (£9 per hour).

“We know that there’s a real drive towards social impact among the AI companies in Cambridge and doing things the right way,” the team explained to me. “And we know that quality of data is important to them, and so we thought we could do a better job of providing quality and ethically sourced data to them, while providing employment to young people.”

They plan to do this by improving the quality of the work they provide to AI companies, with an emphasis on complex tasks, such as those found in medical labeling, and attempt to break these downs into chunks that can be performed by lay people.  It’s an approach that borrows from the success seen in areas such as citizen science, where members of the public contribute to an array of scientific endeavors, often via byte sized chunks of a much larger body of work.

Suffice to say, the very nature of the Hult Prize means that the company are at a very nascent stage of their development, but one organization that is further along their journey are San Francisco-based social enterprise Samasource, who work with gig workers in India, Kenya and Uganda to provide ethical access to the AI marketplace.

Moving people out of poverty

Samasource was founded in 2008 with the express desire to move people out of poverty and into meaningful work.  Whereas most of the workers recruited via platforms such as Mechanical Turk operate as independent contractors, at Samasource, they are full-time employees with salaries and benefits.

“We take people who have not been involved in the formal economy and provide them with the appropriate training and move them into full-time and formal work,” Wendy Gonzalez, Samasource’s President and COO told me recently. “We believe that the right thing to do is to provide people with good quality work rather than aid, and we found people without specific skillsets and trained them to work in an area of good market demand.”

They started in human judgement services such as transcription and curation of data, before rapidly evolving into the provision of training data, which as the AI economy grew rapidly grew to form the majority of their business.

There are currently around 2,800 full-time employees working via the Samasource platform, all being paid a living wage in Kenya, and while this means they are not the cheapest source of labor, the high retention rates and quality of work provided mean they remain competitive in the marketplace.

“We provide our training data at a competitive price, but what makes us different is that we’re not a crowdsourced platform where you don’t know who is at the other end of the line,” Gonzalez says. “We focus in on trust, scale, quality and impact, with all of the people who work in our centers operate from a highly secure building to ensure that the physical and logical infrastructure is secure, and each employees is trained and supported to provide really high quality work.”

Sourcing the right way

Samasource are members of the Global Impact Sourcing Coalition (GISC), which aims to ensure that jobs in this sector are meaningful and sustainable.  This means they have to go beyond insecure gig work and offer people the trappings traditionally associated with a career.

As such employees often have career opportunities within the company. Team leaders and account managers often begin life on the front line, and many of those who don’t move up internally go back to school. Gonzalez says that many employees initially join the company as they couldn’t afford to attend college, so once they’ve saved up some money they return and finish their education.  Typically 80% of employees either return to school or go on to an equal or higher paying job, and they plan to further examine the life chances of employees via a study being undertaken with MIT.

In a bid to stand a better chance against purely for-profit firms, the organization recently shifted from a pure nonprofit to a hybrid structure whereby the founding nonprofit retains the majority of the shares in a for-profit company that can now raise venture funding from investors.  It’s a model they believe can co-exist successfully together.

“I think the profit and non-profit models can co-exist, and what’s really required is the rigor to track and manage the ethical outcomes you want to get,” Gonzalez says. “But there are many other business benefits that come from working like this, whether it’s happy employees, higher retention rates and a desire among employees to become real subject-matter experts in their field.”

While there is still some way before the whole market is as ethical as companies such as Samasource and Meaningful would like, it’s promising to see at least some movement in that direction.  It’s a move that requires going beyond merely seeing this as part of one’s CSR initiatives and towards simply how companies do business in the AI economy.  With formal regulation of the sector not likely to arrive any time soon, it’s beholden on the companies operating in it to push things gradually in a more ethical direction.  Time will tell how effective they prove to be.