Elon Musk and co may relish march of the robots but there must be AI boundaries in the workplace | Heather Stewart
TL;DR
Heather Stewart opens with D4YRL, a robot magician rejected by the Magic Circle: technically competent tricks were not enough without emotional presence and audience connection. The column leans on Sarah O’Connor’s We Are Not Machines, covering Amazon-style worker surveillance, outsourced AI training labor and translators reduced to low-paid post-editing. Its core claim is selective adoption: the question is not whether AI can do a job, but which parts of work society should let companies automate.
Nauti's Take
The strongest point is the power question: workplace AI is not a force of nature, but a management choice with winners and losers. The Musk framing gives the column punch, yet the warehouse, care and translation examples carry the argument better.
Good automation removes dangerous, repetitive or physically punishing work. Bad automation turns people into proofreaders, sensors or machine attachments.
Briefingshow
The piece moves the debate from capability to permission: not every task that can be automated should be automated. The real issue is who gets to negotiate how AI enters the workplace, how work is paced and who controls the data. Without bargaining power, productivity easily turns into surveillance or skill erosion.
Sources