If you ever feel you do not belong in the room, John Hennessy’s advice is to teach the subject to someone else. He still walks into rooms where he is out of his depth, and he says it happens all the time.
I recorded this conversation LIVE at Stanford, and it lands at a moment when any student can look up an answer in seconds. How people actually learn is a live question again, and Hennessy has strong views on it.
This episode is proudly sponsored by ASE!
Pretend you’re the teacher
I asked Hennessy about imposter syndrome, and he said he does not feel it. What he does feel is out of his depth. When he chairs a PhD oral today, he is not the expert, and he works to understand what is going on. In the old days he could follow any computer science defense. Now the field is too deep.
He finds it humbling that thousands of undergraduates on campus know more about their own fields than he does.
His fix is simple: “If you really want to learn a subject... teach it to somebody else,” because then you have to understand it. “Pretend you’re the teacher.”
The lazy mind
When I asked Hennessy what he was bad at, he said lazy. He skipped homework and still got an A on the test, which frustrated his teachers.
At a parent-teacher conference, his high school math teacher told his parents, “John has a very fine mind, but he has a lazy mind.” Then she made him a deal. If he promised to work hard, she would let him take algebra and geometry at the same time, so he could reach calculus by senior year.
He said yes. It was the first time anyone gave him a challenge he could really feel, and he calls it “a real turning point.”
Two wickets and a lesson from MIPS
Hennessy calls himself a reluctant entrepreneur. His Stanford team built a chip several times better than one Intel had just finished with 400 engineers, using four graduate students. Companies still passed on it, so he and his co-founders started MIPS.
They knew a lot about chips and almost nothing about business. In his words, they did not know what a gross margin was, and the mistakes they made meant 20% of the company ended up with investors instead of employees.
His advice to young founders: a new technology and a new business model are “two wickets,” and it is much harder to get through both. Pick one, and if you pick both, raise a lot of money. And when someone says they want to be an entrepreneur but has no technology yet, his answer is, “Step one, go invent the technology.”
AI: overestimated and underestimated
I asked Hennessy what we have wrong about AI. His answer has two halves. In his words, we are overestimating its ability to completely replace human function, and underestimating its ability to replace certain human functions.
He thinks the most interesting AI will augment what people do. He also says learning needs what some call “cognitive friction.” In science and engineering, students learn by solving problems themselves, so teachers have to find new ways to keep that struggle in a world where answers are a search away.
He does not think banning AI in the classroom is the answer. In his words, that is like saying we will not teach calculus because it is mathematics. But AI-generated code still has to be read and checked by a person.
Questions I’m still asking
Who gave you a challenge that felt a little too big, and what did it change?
Where do you feel least qualified right now, and who could you teach it to?
Hennessy believes nobody at the end of their life says they should have worked more hours. What are you protecting?
Try this week
Pick one thing you are learning and explain it to someone in ten minutes, out loud, without notes. Wherever you get stuck is the part you have not learned yet. Go back to it, then explain it again.
Keep reading
Expecting to teach enhances learning and organization of knowledge in free recall of text passages, Nestojko and colleagues, Memory & Cognition, 2014. Students who expected to teach a passage recalled more of it than students who expected a test.
Your Brain on ChatGPT: Accumulation of Cognitive Debt, MIT Media Lab study, summarized in a LIRNEasia journal club deck. In a study of 54 students, essays written with ChatGPT showed weaker recall and brain connectivity than essays written unaided.
Overview of Knight-Hennessy Scholars, Stanford. The graduate scholarship program Hennessy mentions at the end of the conversation, built to develop leaders for hard problems.