←back to thread

Learning to Learn

(kevin.the.li)
320 points jklm | 1 comments | 22 Oct 24 00:01 UTC | HN request time: 0.215s | source
Show context
keeptrying ◴[22 Oct 24 04:26 UTC] No.41911188[source]▶
A big hole in this article is that you need to find the very best learning resource there is. This is a must.

Eg: For RL it would be Barto&Sutton book.

Sometimes the best source is not intuitive. Eg: The best way to become a safe driver is to go to performance drivign school - its a bit expensive but they tell you how to sit and stay alert in a car which I have never seen outside of these schools.

One of my most common things nowadays is to ask ChatGPT is to ask to build a curriculum. Creating and understanding what a great curriculum looks like is 20% of the work of understanding a field.

You can LEARN ANYTHING now if you have the time and inclination and elbow grease. Truly nothing is beyond your grasp - NOTHING. Its a magical time.

I'm actually building a tool that will do all this for you and get you started down the learning path faster than what we have now.

And for the curious - the best way to learn medicine is not a textbook. There are solutions out there like Skethcy which work much better for anatomy.

My own learning project - learn Medicine "on the side". It seems ludcirous that we give up the keys to our health to doctors just so we don't have to learn 2 years of courses. Am going to fix that!

replies(7): >>41911226 #>>41911300 #>>41911448 #>>41911837 #>>41913612 #>>41914030 #>>41916683 #
johnnyanmac ◴[22 Oct 24 06:49 UTC] No.41911837[source]▶
>you need to find the very best learning resource there is. This is a must.

I think there's a line around "good enough", unless your goal of course is to be on the road to "become the very best". I think the better metric is making sure you have a accurate resource over a quality one. The 15-20 hour "sprint hard" methodology isn't stopping after that first sprint, just slowing down.

So if you find/can now access a better resource later, just start the sprint again on that. I know from experience (in real time, unfortunately) how easily "find the best resource" can end up becoming "spend weeks collecting resources but not consuming them".

replies(2): >>41912327 #>>41916562 #
keeptrying ◴[22 Oct 24 17:29 UTC] No.41916562[source]▶
By the very fact that the resource is "good enough" implies that it won't have the holistic and beginner-tolerance required of a good instruction set.

A great text will get you upto beinngin in 2 weeks - month. A "good enough" will mean a year if it isn't your primary focus.

Try learning RL from any other text than Barto-Sutton.

replies(1): >>41919097 #
1. johnnyanmac ◴[22 Oct 24 21:51 UTC] No.41919097[source]▶
Well yeah. Time and budget can be limited. The "best" way to learn a language is to travel to a country native to it and immerse yourself, constantly attempting to speak to others. But that may not be accessible in time nor costs to travel and live. The "best way" to learn many topics is to immerse in a community that's a mix of learners or experts. ie. A classroom setting. A place to constantly iterate and quickly correct mistakes. But education is only getting more expensive, and not all communities are equal.

If there's a great beginner textbook out there, that can definitely help. But not every field has its Bart-Sutton easily accessible. But the time it takes doesn't necessarily matter as much as the mastery itself.