flâneur

blog/2024.09.impact.md at main · okhat/blog

github.com · 3,148 words · saved by 6 readers

Grad students often reach out to talk about structuring their research, e.g. how do I do research that makes a difference in the current, rather crowded AI space? Too many feel that long-term projects, proper code releases, and thoughtful benchmarks are not incentivized — or are perhaps things you do quickly and guiltily to then go back to doing 'real' research. This post distills thoughts on impact I've been sharing with folks who ask. Impact takes many forms, and I will focus only on making research impact in AI via open-source work through artifacts like models, systems, frameworks, or benchmarks. Because my goal is partly to refine my own thinking, to document concrete advice, and to gather feedback, I'll make rather terse, non-trivial statements. Please let me know if you disagree; I'll update here if I change my mind. Here are the guidelines: The fifth bullet "tips on growing open-source research" deserves its own, longer post. I may write that next. This is a crucial mental shif

On Impactful AI Research Omar Khattab | September 4th, 2024 (corresponding tweet) Grad students often reach out to talk about structuring their research, e.g. how do I do research that makes a difference in the current, rather crowded AI space? Too many feel that long-term projects, proper code releases, and thoughtful benchmarks are not incentivized — or are perhaps things you do quickly and guiltily to then go back to doing 'real' research. This post distills thoughts on impact I've been sharing with folks who ask. Impact takes many forms, and I will focus only on making research impact in A

saved by

related reading