Expert Immersion Learning Loop
Do enough homework to ask experts questions that make them think
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 6
- Confidence
- 98%
Hillis learns new fields by combining targeted reading with sustained contact with people who understand the work deeply. He reads enough papers to know the landscape and formulate questions, because approaching someone such as Marvin Minsky or Richard Feynman unprepared wastes their time. The outsider then looks for large gaps or assumptions that insiders may no longer question. Some questions will be naive; useful experts explain why. Occasionally a question exposes a genuinely interesting missing fundamental and opens a deeper conversation. The loop continues through observation, contribution, and introductions to other specialists. Its distinctive mechanism is not networking for access: preparation earns a substantive exchange, and close involvement converts expert explanations and real work into understanding.
Origin
Hillis illustrated the method through his entry into MIT's AI Lab and his relationships with Marvin Minsky, Richard Feynman, and David Agus.
Core principles
- 01People can teach context that papers alone cannot
- 02Homework is the price of asking an expert a useful question
- 03Outsiders can notice questions insiders have normalized away
- 04Admitting ignorance creates room to return to fundamentals
- 05Teaching a newcomer can also sharpen the expert's understanding
How to run it
- 1
Find the real practitioners
Identify people who have built, discovered, or directly practiced what you want to learn.
Pro tip Favor playful experts who remain skeptical about their own field's settled assumptions.
Watch out Prominence alone does not guarantee openness or depth.
- 2
Map the field
Read enough proposals, papers, and foundational material to understand current priorities and terminology.
Pro tip Search for what practitioners say they cannot yet do.
Watch out Do not confuse preparatory reading with mastery.
- 3
Form a live question
Choose a gap or foundational issue whose answer is not already obvious from the material.
Pro tip An outsider's perspective can expose questions a field has stopped asking.
Watch out A question designed only to display knowledge wastes the exchange.
- 4
Earn the conversation
Bring a useful idea, prototype, correction, or thoughtful question that gives the expert a reason to engage.
Pro tip Concrete contribution is stronger than a generic request for mentorship.
Watch out Respect the expert's time and accept a quick rejection.
- 5
Learn through proximity
Observe the work, help solve real problems, and let the expert return to fundamentals while explaining.
Pro tip Take on practical tasks that keep you near the work.
Watch out Passive socializing is not the same as learning.
- 6
Follow the expert graph
Use each relationship to identify other people with complementary or opposing knowledge, then repeat the loop.
Pro tip Seek disagreement as well as consensus.
Watch out Do not outsource your judgment to a single mentor.
In the wild
Hillis read the lab's funding proposals and found an unsolved goal: enabling children who could not read or write to program computers. He built a picture-based method, showed it to Seymour Papert, earned a job and building key, then found and fixed an error in Marvin Minsky's computer diagrams.
→ He gained access to the lab, began working with Minsky, and eventually received a job from him.
Oncologist David Agus brought Hillis a difficult cancer problem. Hillis first learned from Agus, then followed introductions to other specialists while contributing the perspective of a blank slate asking basic questions.
→ Their collaboration developed a protein-centered way to observe cancer processes dynamically.
Common mistakes
Arriving without homework
Without preparation, questions are more likely to waste an expert's limited time.
Mistaking access for learning
Social proximity produces little unless it includes questions, contribution, observation, and real work.
Hiding naive questions
The method depends on exposing missing fundamentals, even though some questions will be corrected as naive.
Is it for you?
Best for
Curious generalists entering a field where tacit knowledge and expert context matter.
Not ideal for
Learners unwilling to prepare, tolerate correction, or respect an expert's time.
From the transcript
“find the people that really understand it hang out with it learn it”
“I read enough papers that I have questions because you're wasting the time of a Marvin Minsky or a Richard fan if you don't ask…”
“sometimes your questions are dumb and they explain to you why they're dumb questions but sometimes they're like yeah that's actually a pretty interesting fun…”
From the episode
#782: Legendary Inventor Danny Hillis (Plus Kevin Kelly) — Unorthodox Lessons from 400+ Patents, Solving the Impossible, Real Al vs. “AI”, Hiring Richard Feynman, Working with Steve Jobs, Creating Parallel Computing, and Much More
Danny Hillis