In the age of AI, the people who win are the ones asking better questions.
The exam asked students to prove they had learned the material by providing five good questions, not answers. It was hard because of what he didn't ask. In that course many great students failed and nobody got a perfect score, which is why the speaker has repeated the method again and again.
They said it was easy, then the room went quiet. Not one of the twelve, between them 200 years of experience and the best people in the company, could put together five good questions. Just silence, not hesitation, not discussion.
Not bad answers and not hallucinations. It is a confident answer, even a correct answer, to the wrong question. AI does not check whether you gave it the right question; it replies confidently and certainly, and it might even tell you how brilliant you are.
Harvard and Wharton gave Boston Consulting Group consultants AI. With AI and good questions, the quality of their work went up by 40%. But that same group, when they trusted AI too much without questioning it, made 19% more mistakes. AI is a multiplier both ways.
Paulo Freire put it together: the teacher deposits knowledge in the head of the student, the student stores it, and the student or employee is more valuable the more knowledge they hold. It made sense while knowledge was scarce, so we added exams and certifications and measured people by their balance of knowledge.
Three words: how might we. Take those words and add a challenge. For onboarding: how might we help new employees feel confident, clear and useful in their first 10 days? Then add tension, such as without increasing management time. A weak question gives you something to look at, a good one something to discover.
How might I make this course fit their schedules, and how could I give people time to think. He had instead asked how to make the five-day course so good that nobody could say no, which was the wrong question. The answer was sitting inside the questions he did not ask.
A lot of research has been done on training people to ask better questions and use AI more effectively, and success rates go up dramatically. In that study people got the job done right 78% of the time with the training, against 45% without it, almost double the results.
Kicked out of four different schools, he covered four years of high school at home in three months, got into top schools and dropped out. At 22 he asked how might I introduce other people to psilocybin, ran retreats in Mexico then Jamaica, and became a successful worldwide bat guano supplier.
The five-day program became two and a half days, and is now ten-minute trainings over ten days, each supplying just enough information to ask a good question, which people then apply. On the last run 76% said they were applying what they learned within 10 days; for off-site training, 20% would be a good number.
“Not one of them was able to put together five good questions.”
“What we need is better withdrawals. And a withdrawal is a question.”
“The biggest risk in the age of AI is not bad answers. It's not the hallucinations.”
“The person who wins isn't the one with the fullest head.”
“A weak question gives you something to look at.”
“Today, the Renaissance man has the good questions.”
“One Thing to Teach in the Age of AI | Bryan Cassady | TEDxUniversityofSalford.” https://www.youtube.com/watch?v=LjOl5GpVWe0. Transcript summary by WhipScribe, https://whipscribe.com/transcript-pages/one-thing-to-teach-in-the-age-of-ai-bryan-cassady-tedxuniver.html.