I have complained before that AI labour exposure estimates are structured guesses rather than theories. If they are hypotheses, their worth depends on whether they predict, and mostly nobody checks.
Leonora Risse and Elise Stephenson report that on Jobs and Skills Australia’s scores, clerical and administrative work carries the highest automation exposure of any sector, and that sector is more than 70 per cent female. Fifteen of the twenty most exposed occupations are female-dominated, and seventeen of the twenty least exposed are male-dominated.
Jobs and Skills Australia also collects internet job vacancy data. In June, vacancies for personal assistants and bookkeepers were running about 22 per cent below a year earlier, and general clerks about 9 per cent below, against a total workforce only 1.7 per cent weaker. Concreters were up 5.5 per cent, electricians up 14.3 per cent.
So a ranking made a prediction and the vacancies moved the same way, in one country, over one year. One agreement does not vindicate the method. What is worth copying is that the authors put their estimate next to data that could have contradicted it.
Assorted links
- Stanford’s Digital Economy Lab, working from ADP payroll records rather than exposure scores, finds a 16 per cent relative employment decline for 22-to-25-year-olds in the most AI-exposed occupations.
- Tencent runs AI camps for thirteen-to-eighteen-year-olds and Geely offers school leavers guaranteed employment at graduate pay, against a McKinsey estimate that China is five million AI workers short by 2030.
- Across 14,419 self-published titles on Amazon, books with sales in a quarter grew 19.2-fold while quarterly revenue grew 8.9-fold, which is what dilution looks like when it arrives as volume.