The Collision – What AI Does to Us

In this episode of Banker’s Bookshelf, Paolo Sironi sits with Eric So, Sloan Distinguished Professor of Global Economics and Behavioral Science, to discuss his new book “The Collision”.

The conversation opens on a distinction most AI talk skips. The industry asks what the technology can do for us. So asks what it is doing to us: to the skills we still bother to build, to our tolerance for difficulty, to what we remember, and to the way we judge the world. He treats AI less as a tool upgrade than as an invasive form of intelligence already reshaping the people who use it. The frame is Steve Jobs’s 1980 line that a computer is a bicycle for the mind, the most efficient mover on the chart. So’s revision is blunt. In the age of AI, the bicycle has become a motorcycle. It offers speed and self-propulsion, and it multiplies the cost of a bad decision. The title is chosen for the same reason. “Adoption” and “shift” sound orderly. “Collision” names a rapid force that is changing human intelligence and machine intelligence at the same time.

From the research, he draws four rules. Value the struggle: products sell the removal of friction, but struggle is how knowledge becomes yours. Students who outsource essays look strong at home and collapse in the exam room. Value who you are without AI: a polished deck you cannot defend is not your work, and the same failure shows up in a client pitch. Reinvest the surplus: history’s transformative tools did not merely speed the old task; they let people reach harder problems. Saved time spent scrolling is a wasted surplus. Make AI a cognitive trainer: personalized tutoring has long delivered a two-sigma learning gain, and AI can approximate it only if it walks you to the answer instead of handing it over.

Banking sits in the middle of the argument. The sector’s instinct is to strip friction and automate cost. So’s warning is cognitive short-termism: fewer juniors, less mentoring, and a thinner bench of people who can still trace a problem when the model is wrong. Efficiency that deskills the institution is not a saving. It is a debt.

The episode is a call to treat AI as a behavioral problem, not only a productivity one.