Worst Ideas for Managing the AI Transition

AI companies face strong incentives to release products, minimize harms and define safety on convenient terms. Voluntary promises cannot replace independent testing, liability and enforceable rules.

Not every displaced worker can or should become a software developer, and coding itself is being automated. Retraining must connect real people to actual regional demand.

Using AI to investigate benefit recipients can scale biased data, opaque suspicion and false accusations against people with little ability to appeal. Administrative efficiency does not justify automated punishment.

Teachers do far more than deliver explanations: they motivate, supervise, notice distress and build a classroom community. Replacing them wholesale would mistake content generation for education.

Requiring constant monitoring in exchange for work, insurance or public services would make privacy a luxury. AI does not need every human action converted into data to be useful.

Cash transfers cannot substitute for functioning healthcare, education, housing and transport where supply is scarce. Cutting services while paying a dividend would leave recipients bidding against one another.

A blanket ban would push important research into less accountable places while blocking defensive, medical and scientific work. Governance needs distinctions between dangerous capability and useful experimentation.

Taxing every automated system identically would punish harmless productivity tools and miss the companies capturing exceptional rents. Policy should respond to labor effects and market power, not the word robot.

Clinicians combine examination, history, consent and responsibility in situations where information is incomplete. AI can assist diagnosis without becoming the accountable relationship through which care happens.

A race with no standards rewards the company or country most willing to ignore safety and labor costs. Shared testing and reporting rules can preserve competition without making recklessness an advantage.