Highly skilled workers, who are crucial in training artificial intelligence systems, face significant risks as their expertise becomes digitized. While their contributions enhance productivity and strengthen organizations, once their knowledge is encoded in software, it no longer exclusively belongs to them, and they are often not compensated extra for providing it. In some cases, this could lead to higher-skilled workers being replaced by less-skilled workers, who are then supported by the very AI models that the former group helped train.
AI's ability to digitize expertise allows companies to scale an individual's skills globally and across time. For instance, in a call center study, an AI assistant, trained on transcripts of top agents' conversations, significantly improved problem resolution, especially for newer workers. This leads to less experienced agents efficiently resolving issues and increased customer satisfaction. Coding tools like GitHub Copilot have also shown to help users, particularly junior developers, complete tasks faster.
However, this transformation introduces new competitive pressures for knowledge workers. While they might perceive their skills as unique within a company or region, data makes the market for expertise global. Anyone capable of generating similar outputs can provide the data needed to build an AI model that can replicate their work. This exposes individuals to new competition from various locations, potentially weakening their collective bargaining power unless workers coordinate.
A recent FT-Focaldata poll revealed a widening divide in AI adoption, with over 60 percent of the highest-paid workers in the US and UK using AI daily, compared to just 16 percent of the lowest earners. This disparity, also evident in a persistent gender gap where men are more likely to use AI tools, suggests that AI could exacerbate earnings inequality. Corporate training is identified as the biggest driver of AI uptake, and experts like Nobel laureates Daron Acemoglu and Chris Pissarides predict that AI will increase inequality between labor and capital, and that IQ will become increasingly important. The question remains how quickly structured AI training can be scaled to close these gaps, a key policy consideration in the UK resultsense.com.
