Martin Remy

The strength of “I don’t know”

I had a professor in grad school who sometimes threw erasers at students. Paul Halmos was from that generation of reluctant Hungarian emigres who made big contributions to science and computing in the US during and after World War II (and, yes, atomic weapons – although Halmos didn’t work on the Manhattan project as far as I know). I was lucky to have taken classes with him late in his career.

Besides affectionately pelting students with erasers, Halmos had another, more endearing, habit.  He would write a theorem on the board and then turn to the class and say, “How do we prove this? I have no idea! Let’s get started!”  Of course he knew the proof, but by saying “I have no idea”, he got everyone’s attention and turned us from passive scribblers to active participants.

If you’re shipping AI products, you don’t have time for weeks or months of discovery, PRDs, and “planning.”  But make sure your team is aligned and that you’ve heard from the smartest people in your org before you commit and go. One meeting or an async thread lasting a day or two can keep you from wasting time on a bad idea.

  • “I don’t know” starts a discussion and opens the floor for other perspectives.
  • “I don’t know” welcomes dissent and defeats a culture of fear (of disagreeing with leaders).
  • “I don’t know” enlarges the solution space. There might be a much better that’s adjacent to yours.
  • “I don’t know” creates space for the quiet, intelligent people in the room to contribute. You might hear a key new insight that unlocks a breakthrough for the project.

Photo of Paul Halmos by George Bergman from Wikipedia

2 responses to “The strength of “I don’t know””

  1. Alex Bendig Avatar
    Alex Bendig

    Leaning into not knowing, not being certain is so valuable, but also pretty counterintuitive, I think. Per Daniel Kahneman, “A remarkable aspect of your mental life is that you are rarely stumped.

    Since you are bringing up AI products – they actually reflect that bias beautifully. They seem to generally insist on having an answer, rarely if ever admitting not knowing.

    Liked by 1 person

    Reply
    1. Martin Remy Avatar
      Martin Remy

      Per Daniel Kahneman, “A remarkable aspect of your mental life is that you are rarely stumped.

      Yeah, we can usually spit out an answer, like LLMs (with or without reasoning). I think the value I’m trying to articulate is two-fold:

      1. Sometimes expressing uncertainty as a way to signal openness to others’ opinions is very helpful (even for LLMs, but it is rare unless prompted). Even if I’m not stumped, but I have a vague sense that I might be missing something potentially important about a decision.

      2. Sometimes fabricating uncertainty for yourself is helpful: “How might I be wrong about this?” is a question people should ask themselves more often. Not as comprehensive or morbid as a full pre-mortem, but as a way to shake us out of that un-stump-able rut.

      Like

      Reply

Leave a comment