This past week I discussed learning strategies with my brother-in-law who is currently in university. Here’s a summary:
Learning shouldn’t be easy.
Karpathy has an amazing tweet about learning. I highly recommend reading it for yourself. The most important piece says:
Learning is not supposed to be fun. It doesn’t have to be actively not fun either, but the primary feeling should be that of effort. It should look a lot less like that “10 minute full body” workout from your local digital media creator and a lot more like a serious session at the gym. You want the mental equivalent of sweating. It’s not that the quickie doesn’t do anything, it’s just that it is wildly suboptimal if you actually care to learn.
Simulating this is difficult, especially in the age of AI, where an answer can be given immediately. However, for topics you want to learn deeply, taking the immediate answer means you are most likely short-changing yourself. AI is great at being a world-class tutor. All it takes is prompting it the right way. Recently, I have found this prompt to be extremely helpful when learning something new:
I'm learning [TOPIC] and I want you to act as a demanding tutor, not an explainer. Here's my current understanding: [PASTE A ROUGH PARAGRAPH OF WHAT YOU THINK YOU KNOW]. Don't give me answers or summaries. Instead, ask me pointed questions that expose where my understanding is shaky, make me reason out loud, and push back when my explanations are vague or wrong. Only give me a hint after I've made a real attempt, and keep it minimal so the work stays mine. Don't move on to the next concept until I can explain the current one to you in my own words without errors. When we finish, tell me the two weakest spots in my understanding and what I should go read to fix them.
Using this prompt with your favorite agent / AI is really beneficial for a few reasons:
- The AI / Agent can identify your gaps in knowledge (and help you close them).
- It forces you to have the uncomfortable learning (which will help you remember).
- It’s a much more efficient and dense use of time compared to just reading slides 100 times.
The goal is that the agent is pulling you along and gives you just enough help that you barely overcome the obstacle. This way, it will be something you never forget.
Practice game reps.
Once you’ve gone through the process above, you should be able to quickly identify and execute the learning process for virtually any topic. Once you feel you have a good base of knowledge, it is important to practice game reps. Just as professional athletes practice full speed in a game-like environment, it’s important for you to do that as well.
Why would we prepare for tests and interviews differently than what the test or interview is actually like? The super power of AI is you can create these game-like environments easily. Once you have a topic, you can ask the agent to generate as many “real tests” as you desire. You can then take the tests under the same conditions you would take the actual exam in. It forces your body and mind to get accustomed to the stress, fatigue, and environment of the actual test.
If I had a test in 5 days I was preparing for, I would have an agent generate two tests a day and take each test with pen and paper like it was the real deal. This would give you ~8-10 tests that you’ve taken before you show up. When you find areas in the test that you struggle, you can revisit the learning process to close the gap.
Alternative Learning Formats
Finally, having your agent provide alternative learning formats is a great way to see things in a different way. Sometimes it takes a different view or format to really get you over the hump. In engineering school, I would have loved to have different demos, games, and visual diagrams for complex topics.
For example, here’s a skydiving game for learning velocity and acceleration. Air resistance grows until it matches your weight. Watch how opening the parachute changes your acceleration before it slows your fall, and try to land gently:
- Altitude
- 800 m
- Velocity
- 0.0 m/s
- Acceleration
- 0.00 m/s²
Jump, then pull your parachute as low as you dare. Land slower than 7 m/s.
If I had to sum up what I told my brother-in-law: use AI to question you, test you, and show you things a different way. The sweat is where the learning happens.