i have a monthly offline test for the subject "Machine Learning Techniques" for the IITM BS course tommorrow. the syllabus covers topics taught over 4 weeks. for the first 2 weeks, i was unprepared because the topics didn't connect - linear, ridge, lasso and kernel regression. i'm not so great with matrices and the sight of (XXT)^-1 scares me.

parkinson's law took over during these last 2 days because i had to learn no matter how. 

i opened up PYQs and started solving them. those that I couldn't solve i copy pasted to ChatGPT and asked it to solve as well as explain the relevant concepts. this is backward learning. generally what we are taught with is forward learning. learn the concepts, do practice questions then attempt tests. 

this method goes straight to test questions, and you try to solve them. at first you can't so you read concepts relevant to solving the question. then repeat. 

and for some reason this works so effectively. i'm positively surprised. i didn't think i'd have been able to do the topics in the last 2 days but i'm much more confident now. 

AI is now good enough for personalised tutoring upto the level of masters degrees. what remains is societal level adoption and diffusion. 

personal tutoring + neuroscientifically effective ways to learn can form strong synapses in brains fast.