Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Anki and GPT-3

andrewjudson.com

Post reply on HN
Print view View original

21–30 of 80 posts

Re: Anki and GPT-3

#21

Post by DecayingOrganic »

Version467 wrote: post #10

As a daily Anki user, I am highly skeptical of this idea. The careful creation of new cards is one of the key parts in learning the material. It promotes a critical thinking step where you have to distill the material down into small but important pieces and (if you want to do it well) forces you to evaluate if you have actually understood what you wrote down. I suspect that forgoing this step will not lead to unders…

Anki's key benefits stem from its utilization of two highly supported and extensively researched pedagogical strategies: (1) retrieval practice and (2) spaced repetition. In contrast, the practice of note-taking, which essentially entails summarizing information in one's own words, has been found to have 'low utility' in academic literature. [0]

[0]: https://sci-hub.ru/https://pubmed.ncbi.nlm.nih.gov/26173288/

Re: Anki and GPT-3

#22

Post by grandpoobah »

rjh29 wrote: post #14
Version467 wrote: post #10

As a daily Anki user, I am highly skeptical of this idea. The careful creation of new cards is one of the key parts in learning the material. It promotes a critical thinking step where you have to distill the material down into small but important pieces and (if you want to do it well) forces you to evaluate if you have actually understood what you wrote down. I suspect that forgoing this step will not lead to unders…

It's a classic beginner mistake to load up a 20000 word deck and start learning it when you have no relationship or connection to the cards. Create a card when you first see the word in real life, with that example sentence on the back, and it's so much easier to remember. At least for me anyway.

Those large word decks can be useful though. Download deck, immediately suspend all cards, and then start un-suspending cards as you learn the words. I also add notes to the cards as I go.

I have a similar strategy for the Ultimate Geography deck. I want to know where every country is on a map, as well as their capitol and flag. So I suspended all cards, and then periodically enable 20 new ones, and learn as I go. That often means googling the location or pronunciation the first time I encounter the card.

(I use the low-key Anki configuration with Pass/Fail buttons plugin)

Re: Anki and GPT-3

#23

Post by coffeeri »

IanCal wrote: post #16

This is pretty cool. One of the things I think is hugely important about GPT3 is that it's paid for by the token. So for small users, it's pretty cheap, and it's fully workable for testing and experimentation.

This statement is accurate for testing and experimentation purposes. Additionally, the $18 starting credit provided by OpenAI is a nice bonus.

I created a small Anki plugin that tags cards based on their content using the Text-Davinci-003 model. However, the cost quickly becomes prohibitive when a large number of tokens/ Anki cards need to be processed. To be fair that plugin is not working great, since tags should be viewed, merged and reused on a global scope.

Re: Anki and GPT-3

#24

Post by rjh29 »

grandpoobah wrote: post #22
rjh29 wrote: post #14

Earlier quoted context omitted.

It's a classic beginner mistake to load up a 20000 word deck and start learning it when you have no relationship or connection to the cards. Create a card when you first see the word in real life, with that example sentence on the back, and it's so much easier to remember. At least for me anyway.

Those large word decks can be useful though. Download deck, immediately suspend all cards, and then start un-suspending cards as you learn the words. I also add notes to the cards as I go. I have a similar strategy for the Ultimate Geography deck. I want to know where every country is on a map, as well as their capitol and flag. So I suspended all cards, and then periodically enable 20 new ones, and learn as I go. Th…

Everyone's different, but the act of creating the card itself helps motivate me to learn it. I really enjoy having a hand-curated deck. Though I can see larger decks being useful if the card is annoying to make (like your geography example).

Re: Anki and GPT-3

#25

Post by IanCal »

coffeeri wrote: post #23
IanCal wrote: post #16

This is pretty cool. One of the things I think is hugely important about GPT3 is that it's paid for by the token. So for small users, it's pretty cheap, and it's fully workable for testing and experimentation.

This statement is accurate for testing and experimentation purposes. Additionally, the $18 starting credit provided by OpenAI is a nice bonus. I created a small Anki plugin that tags cards based on their content using the Text-Davinci-003 model. However, the cost quickly becomes prohibitive when a large number of tokens/ Anki cards need to be processed. To be fair that plugin is not working great, since tags should b…

At 2 cents per 1000 tokens (~750 words), unless you're doing it a lot that should be quite manageable? That's ~$13 for 1000 cards of 500 words (prompt + card + tags). The smaller models are cheaper still and might perform well enough.

Re: Anki and GPT-3

#26

Post by yummytummy »

Very cool! I've experimented with a similar thing for trivia questions based on Wikipedia articles.

I'm also trying to reference the part of the text where the question was generated from, so the user can verify if it was correct or learn more about it in context.

https://trivai.app/

I currently require a sign up to generate questions, but the latest questions can be seen at

https://trivai.app/latest-questions

Re: Anki and GPT-3

#27

Post by duwip »

Silly question, but how does one get an Anki API Key? I've installed anki-connect and can't quite figure out where to get the key...

Re: Anki and GPT-3

#28

Post by JonathanFly »

DecayingOrganic wrote: post #21
Version467 wrote: post #10

As a daily Anki user, I am highly skeptical of this idea. The careful creation of new cards is one of the key parts in learning the material. It promotes a critical thinking step where you have to distill the material down into small but important pieces and (if you want to do it well) forces you to evaluate if you have actually understood what you wrote down. I suspect that forgoing this step will not lead to unders…

Anki's key benefits stem from its utilization of two highly supported and extensively researched pedagogical strategies: (1) retrieval practice and (2) spaced repetition. In contrast, the practice of note-taking, which essentially entails summarizing information in one's own words, has been found to have 'low utility' in academic literature. [0] [0]: https://sci-hub.ru/https://pubmed.ncbi.nlm.nih.gov/26173288/

According to some people (Justin Sung on YouTube is where I heard it, probably, so no idea if credible) you can influence the forgetting-curve and make Anki more effective by having more context, putting knowledge into relationships with other knowledge, etc. It's a multiplier on raw/rote space repetition.

Re: Anki and GPT-3

#29

Post by mclightning »

Version467 wrote: post #10

As a daily Anki user, I am highly skeptical of this idea. The careful creation of new cards is one of the key parts in learning the material. It promotes a critical thinking step where you have to distill the material down into small but important pieces and (if you want to do it well) forces you to evaluate if you have actually understood what you wrote down. I suspect that forgoing this step will not lead to unders…

Not even factual trivia

Re: Anki and GPT-3

#30

Post by oaktrout »

Version467 wrote: post #10

As a daily Anki user, I am highly skeptical of this idea. The careful creation of new cards is one of the key parts in learning the material. It promotes a critical thinking step where you have to distill the material down into small but important pieces and (if you want to do it well) forces you to evaluate if you have actually understood what you wrote down. I suspect that forgoing this step will not lead to unders…

Creating cards is great, but for those seriously deep in the anki game (eg 20,000 active cards or 20 new cards per day for years straight), there isn't enough time to create all the cards you need to learn and pre-made decks are king.

I imagine if you need to learn that much material but the pre-made deck isn't there, this gpt integration might be very helpful.

Post reply on HN