31–40 of 154 posts
Knowing things is really important, but I don't think learning makes you "smart"; it just makes you know more things.
A hacker is just someone who has practiced learning independently and has become exceptionally good at it. The reason people say you can't teach the hacker mindset is because without the underlying drive, there's nothing you can tell anyone. It's like when teachers who lament students don't care what they say so long as they get the right grade, it's because those students are optimizing for approval in a system because that's sufficent for their limited purposes. The more you profess to them, the more you reinforce that learning is passive submission to authority. If you want to make hackers, start with necessity, and technique will emerge as the artifact of navigating constraints. If you want to make people smart, challenge them instead of just telling them things. Hackers aren't defined by knowing more, they're defined by having physically done more. Spaced repetition as it's usually presented optimizes for outcomes in an approval environment that produces people who have been rewarded for cheating themselves out of knowledge and expereince.
I would say, want to learn physics? Build mechanisms or make radios. Number theory? Break cryptosystems. Astronomy and geometry? Sail at night. Lead? Ride horses. Fluid dynamics? Tune engines. Statistics? Write a spam filter. Speak a language? Tell their jokes, etc. Imo, most education is set around meaningless but scalable exercises of professed skills instead of meaningful exercises that are more powerful, but don't scale. We've optimized for scale at the expense of quality. It's the solution to an inferior problem.
So sure, learn spaced repetition, but really, find something and practice it for more joy and better results instead.
I did want to give some feedback, though. I think this post suffers from too much hand-waving, which is what plagues most other posts about learning and spaced repetition (excluding probably just Gwern). For example, it compares flash cards and reading to just reading, citing the results of the study as a negative:
> This is however a very slow process. One study implies that in the best case it takes adults 10 years of reading 1 hour a day to get twice as effective at reading. Even if this is technically learning exponentially, the improvement rate is so slow that the process is indistinguishable from a linear one.
...
> We said in the best case it takes adults 10 years to get twice as effective reading. With spaced repetition it takes only days for your time to get twice as effective. These growth rates are completely different.
Maybe the rate of change of effectiveness of the reading is slow, but does that matter if you're accumulating knowledge from all of that reading, especially as it builds off of prior knowledge? I also don't think it's a 1:1 comparison to contrast these. It only takes days to get twice as effective at reading with spaced repetition? Or at learning? If it's the latter, I don't think that's what the earlier study measured?
The other thing that jumped out at me is the huge focus on spaced repetition and memory for learning, which are absolutely helpful, but there seems to be a lack of what constitutes memorizing versus understanding (and I'm not sure I see that in the Learning is Remembering post either). I think about other ways to build your understanding, like working through problems and applying the knowledge, that are key to learning. Much of learning physics is getting your hands dirty in the equations, and there's a big difference between knowing a formula and really understanding it in action.
It would be great if someone could provide an example to an old dummy like me.
Earlier quoted context omitted.
Is this what you are looking for? https://ankiweb.net/shared/decks/
I find the Anki shared decks are basically not high quality enough. Do you ever use them? Which ones do you use?
Would very much like a version control system for Anki decks so updating cards can be done by the community and transparently. This would allow people to validate and properly maintain decks that change with time (like programming languages / regulatory requirements / operating standards etc).
Spaced repetition turns reading into a physical activity with feedback. If you had an activity that required the knowledge you were hoping to acquire, you would learn at the same rate or faster by practice. Most lessons are encoded really poorly. I think spaced repetition is great, but practice at something is better. A hacker is just someone who has practiced learning independently and has become exceptionally good…
Spaced repetition turns reading into a physical activity with feedback. If you had an activity that required the knowledge you were hoping to acquire, you would learn at the same rate or faster by practice. Most lessons are encoded really poorly. I think spaced repetition is great, but practice at something is better. A hacker is just someone who has practiced learning independently and has become exceptionally good…
Also learning a new (natural) language sometimes requires hammering words or gramatic rules into memory, and having a good teacher can be much faster than learning on your own by reading texts.
My idea is that there is a market for community-curated spaced repetition decks. Many people want to learn the same things such as a foreign language or a programming language. The difficult part is creating a deck and crafting the answers and questions. Because usually this is a time-consuming process. So if it was a community-effort then it would be a win-win. This is probably not an original idea and if anyone kno…
For language learning, there are flash card generators that make this a simple one-click process. I think these strike the right balance of simplicity & flexibility/personalization. Of course, a tool like this relies on a free database that you can map concepts onto. But I could see this sort of working with wikipedia or some documentation.
Edit: to the article's point spaced repetition to memorize domain specific facts is useful but it's not exponential like reading is.