The AI pay boost could peak sooner than you think, professor says (www.businessinsider.com)

Penn professor Ioana Marinescu, coauthoring a Brookings paper with Konrad Kording on “intelligence saturation,” warns that the initial AI wage boost may already be approaching its peak. Using a model that maps wages onto the share of cognitive (“intelligence”) tasks automated, she estimates the economy has automated roughly 14% of intelligence tasks so far and identifies a potential tipping point around 37% automation where wage growth could reverse. The early phase—where AI augments productivity and raises pay—can give way to declining wages if automation replaces enough intelligence-sector jobs without new, high-value roles emerging. Key technical points: the model assumes easiest-to-automate intelligence tasks (pattern recognition, translation, basic writing) are displaced first, and the timing of a wage downturn depends on parameters like substitutability between intelligence and physical work and substitutability within intelligence tasks. Empirical signs to watch are slowing wage growth and a falling share of workers in intelligence-heavy roles (Marinescu cites a decline in routine cognitive job share from 49% in the late 1970s–80s to 35% in 2018 as a proxy for ~14% automation). If software developers, translators, or copywriters are lost without replacement by new cognitive roles, the AI pay curve could flatten and then decline—making monitoring sectoral employment shares the most important near-term metric.