Item id: 49095106
2 hours ago
Actually you should not compete in "AI Study Assistnat"(just answer questions), you shouold better do something like use prompt design to constrain AI's output, and let AI to analyze user's level or trends. (sth like adaptive learning) good prompts can prevent overfitting
9 hours ago
This sounds kind of like ChatGPT Study Mode.
16 hours ago
If you want to create an AI that will be better at _teaching_ and not just answering questions, you may want to talk to a human teacher about their process.
"Pedagogy" is the term for professional teaching; "pedagogical methods" would be a good term to use in searches, or in querying AI, about processes like this.
14 hours ago
The current LLMs can show their work, if you don't understand it, ask to have it explained it in a different way -- idk how you can improve on that experience.
4 hours ago
You say, in much fewer words, what I was writing so will just +1 you. Maybe the Chat GPT Study Mode mentioned above does this functionality already; would be surprised if this isn't already available.
17 hours ago
Output should be a various ways to solve a problem right?
Im ignorant about mathematic domain in AI, but if there arent the ones that trained specifically tuned for mathematic domain then you should fine tune the open model.
Also, if you are not asking for exact answer, gpt or claude can still give you an insight how to solve it.
Having thought that if chegg couldnt build one same as your idea, then it would not be easy cause they have tons of data already