Artificial Intelligence used in workplaces is frequently “ineffective, simplistic and opaque”, often making human workers’ jobs harder and relying on cheap human labour from the Global South to “automate” processes, according to a new study by British academics.
A study by the Oxford Internet Institute, AI @ Work [pdf], analysed over 400 academic and industry reports between 2019 and 2020 to find the outstanding challenges of the emerging technology and how big the gap between AI technologies and the environments where people use them is.
The report authors, Gina Neff, Maggie McGrath, and Nayana Prakesh, say the gap is considerable, especially “in how AI tools [are] used and how people talk about what they are supposed to do”.
The authors say there are three broad ways AI fails workplaces: integration, over or under reliance on AI technology, and transparency – when AI is being used and when it is not.
“Ultimately serious gaps remain between the social and technical infrastructures required for functional AI in many workplace settings,” said University of Oxford Associate Professor and Senior Research Fellow, Gina Neff,
“Until these gaps of integration, transparency and reliance can be solved by workers in their workplaces, AI tools and technologies will continue to demonstrate serious—and sometimes dangerous—shortcomings in practice.”
According to the report, the cost and legwork of AI workplace projects means it is still difficult to justify despite billions of dollars of investment in recent years and optimistic claims from vendors.
The report says often the majority of work involved in AI projects is in data collection and cleansing, and companies rely on the Global South where labour costs are lower, but market their solutions as automated or ‘AI enabled”.
“AI is marketed as a solution to many problems, from healthcare diagnosis to replacing white-collar workers. However, much of the work that is attributed to AI is currently still done by human labourers —often in countries such as India and China where labour costs are lower.
“Weak data privacy laws and cheap labour in other countries mean the time-consuming work of data cleansing and data cataloguing can be outsourced overseas.”
An example cited in the report is Indian startup Engineer.ai, which claimed to use AI to let users build apps in around an hour. But an investigation by The Wall Street Journal uncovered the company, which attracted nearly $US 30 million in funding from a SoftBank-owned firm and others, was actually relying on cheap Indian engineers to build apps.
The Oxford Internet Institute report says, “These cases show that AI is marketed as automatic, but often based on tedious behind-the-scene work … While the phrase AI attracts funds, the work in building and ensuring that the systems work is relegated to lower paid workers.”
The report also found AI can make workers jobs harder, forcing creative workers, for example, to wrangle with algorithms to ensure their work is seen and can generate revenue. There is also evidence of workers being used to train AI algorithms to make workplaces more efficient and being under increased workplace surveillance through AI tools.