The Job Apocalypse Is Looking Less Like a Wave

The fear that AI will arrive as a sudden, sector-wide redundancy event has been a fixture of policy debate for three years. New work from MIT's Computer Science and Artificial Intelligence Laboratory and FutureTech offers a useful corrective. Their preprint, "Crashing Waves vs. Rising Tides," argues that AI is more accurately understood as a slow-moving, broad influence on work than a discontinuous shock that wipes out specific jobs.

The methodology

The methodology is unusually grounded. The researchers identified 11,500 tasks in the U.S. Labor Department's O*NET database, generated multiple instances of each, and ran them through more than forty AI models with realistic workplace prompts. They then asked workers from those fields to evaluate over 17,000 model outputs and judge whether the work was good enough to use without edits.

The headline numbers: in 2024, AI could complete around 50 percent of text-based tasks at a minimally acceptable level. By 2025 that figure had risen to 65 percent. At the current pace, the projection reaches 80 to 95 percent by 2029.

Reshape, not replace

What does that mean for jobs? Mostly that they change rather than vanish. As MIT Sloan summarised the related findings, task automation does not equal job loss in most cases. Roles get redrawn around what AI handles well, and humans are pushed toward judgement-heavy or relational work that the models still botch. Axios's coverage framed the result as a direct challenge to the AI-job-apocalypse narrative.

The honest caveats

This is not a clean win for AI optimists. Senior author Neil Thompson is careful to note that "tides could still rise quickly" and that gradualism is not inherently protective. Workers in highly exposed jobs still face real pressure on wages and hours. There is also early evidence that companies are laying off staff in anticipation of AI's promised productivity gains rather than in response to demonstrated ones, which is a different kind of harm but a real one.

The myth worth retiring is the one where 30 percent of US labour disappears in a quarter because GPT-X arrived. The reality, as the MIT data describes it, is closer to a long reshuffling of which tasks belong to humans and which belong to machines. That is still a serious adjustment for workers caught in the wrong categories. It is not the cliff most of the popular discourse implies.