ChatGPT is great at taking my notes and turning them into Anki cards. Some tweaks are occasionally needed, but I have done well on numerous exams letting ChatGPT figure out the cards (and the formatting tedium) for me.
Would you mind giving an example of the kind of prompt(s) you use to get these card generated?
On a different note, ChatGPT has been making me feel stupid because I fail to come up with many use cases, while it seems like tons of technically unskilled people come up with all kinds of uses...
I'm no good at prompt engineering, so I iterate a lot, rather than trying to come up with the perfect prompt. I start by asking something like,
"please generate a series of short questions and answers that conform to a flashcard format, easy for studying, from these notes"
The responses were too verbose, so I asked:
"Please shorten the answers to short, easily memorizable lines"
That returned pretty good flashcards, so from there I asked it to reformat in a way Anki could consume:
"these are great, please convert these into a code block, where the question and the answer is on the same line, separated by a semicolon and not numbered"
But that put all of the flashcards on the same line, so I added:
"sorry I wasn't clear, each question should be on its own line"
That gave me what I was looking for. The next step is to paste it into Anki, fiddle with the recall and cards-per-day settings, and then get to training!
If you’re modifying the cards to fit your needs and ChatGPT is basing them off of your personal notes then I’m not sure you understood the commenters suggestion you’re referring to.
One of the biggest roadblocks to using anki or cards in general is coming up with good cards. Automating that part lets you use anki a lot more, which is a win. If you want to connect with the question you can still manually transcribe the card.
The biggest failure mode is adding to many notes at once, which leads to non obvious increased future workload. Automating card creation sounds like would worsen this issue unless extreme care is taken.
Using Anki is not a goal in itself. It's just a tool to help you cram facts into your brain. Once you have done that, you can and should build associations and knowledge on top of them. Reviews ensure that less-often used facts are not forgotten. Anki is fine for that.
But creating Anki cards is effectively a form of review as well. If you automate that, then you miss out on going through your own notes, engaging with the subject, and subdividing it into self-contained pieces, which is required to come up with good cards.
I would rather get cards that are really well designed ___as cards__ (which chat gpt will do better than me given the time I want to spend) even if there is some loss of "value" from not breaking the material into cards myself, because I will make it up on the back end by studying the cards more, and having cards that are more correctly designed for optimal study. Who cares if I did the work of breaking the material into cards myself if it results in shitty cards that aren't good study aids and thus you never use.
Why would you end up with shitty cards if you did them yourself? And how can you even tell whether ChatGPT made good ones? And how much of a difference do the "better" cards even make?
There is not necessarily a tradeoff here. Sure, you will make some errors, but if you find a card to be so bad that it's useless for SRS (unlikely, unless the answer is very vague or too long to quickly validate) you will just delete it and replace it with a better one.
There is a "skill" to everything, and my card making skill is probably a 3. It takes me some time to identify nuggets that will be long enough not to be trivial but also short enough to make good cards and salient enough to bother with. ChatGPT on the other hand is probably a 7.5-8 in making cards, and it can crank them out. On top of that it can suggest things that you are probably going to be tested on given the stuff in your notes that you might miss if you just created cards from the contents of your notes. If you have a famous teacher at a public university who teaches every semester for a big core class, it can even go far beyond that in ways that will surprise you.
Of course you don't have to create your own cards, but they would work better if you did. You can use ChatGPT to automate onerous tasks, but have to compensate for it by engaging with the material in other ways. I doubt that grinding decks is enough. Depends on the activity and the goal I guess. Acing exams, sure, without doubt.
A concept applies here where you really don't want to spend unbound time studying. It's a great way to burn out. "A" students, when interviewed, frequently reference doing "enough" and then stopping.
Optimizing your study time by taking the "boring" work, e.g. building flashcards of material you organized yourself, cuts down on time spent not doing work that's maximally effective (memorizing flashcards).
If a person had infinite time, making flashcards would be just one more thing on an unbound list of things they could do to learn the material better. But people don't have infinite time, and considering you're already "familiar" with the material since you wrote and organized it, building flashcards is something you can't really afford to spend time on.
When I went back to school to get my undergrad, I focused on optimizing my time, because procrastination was a huge problem for me the first time around. Being efficient and not spending a single second longer than I have to in order to get an "A" is important to me, and using ChatGPT to help me with that has been wildly successful so far.
I find in my usage that creating a card can take several minutes (5-10 for a hard card) but reviewing them takes a few seconds per occurrence. To reach the initial creation time would take me decades of reviews.
That said, I still make all my cards myself and some cards are trivial. For the trivial cards, review time can outweigh creation time. But trivial cards take seconds.
Crafting effective prompts is important to getting the value out of an SRS. That sometimes is the bull of the "work". Not the review.
