How do you find the time to read soany papers and how do you self-evaluate if you have absorbed the paper enough (surely the pareto efficiency lies somewhere in between reading everything implementing nothing and reading little and implementing everything)
running a weekly reading group (https://aman.ai/primers/ai/top-30-papers/, 4-5 papers/week) really helped bootstrap! i currently implement very little. would like to change that but i find it hard to estimate how long implementation will take & how worth it it will be, vs. evaluating how the paper connects to extant research questions/priorities & generating new research directions based off it.
would be interested in how you evaluate what to implement tho writing up probs not best use of ur time
Hey, great read as always. Your discussion on explicitly training models for generalization is quite timely; I often think about how diverse movement patterns in Pilates contribute to a more universally adaptable physical form. Do you anticipate future AI architechtures will intrinsically prioritize this broader adaptability?
hmm now I want to do one as well. Good update.
How do you find the time to read soany papers and how do you self-evaluate if you have absorbed the paper enough (surely the pareto efficiency lies somewhere in between reading everything implementing nothing and reading little and implementing everything)
running a weekly reading group (https://aman.ai/primers/ai/top-30-papers/, 4-5 papers/week) really helped bootstrap! i currently implement very little. would like to change that but i find it hard to estimate how long implementation will take & how worth it it will be, vs. evaluating how the paper connects to extant research questions/priorities & generating new research directions based off it.
would be interested in how you evaluate what to implement tho writing up probs not best use of ur time
Hey, great read as always. Your discussion on explicitly training models for generalization is quite timely; I often think about how diverse movement patterns in Pilates contribute to a more universally adaptable physical form. Do you anticipate future AI architechtures will intrinsically prioritize this broader adaptability?