mastodon
4.6.6
I like to write code
My sports and visualization interests live mostly on bluesky https://bsky.app/profile/billmill.org which you can follow on mastodon as @billmill.org
Portland, Maine, USA
You're absolutely right! You're absoltuley right! Yoou're absolutely rright!ｙou're absoluttе үrihht! You'er abs૦Ɩtuеⅼｙ riigtһ! уou'rrе bsｏυetlүy rgι!t уuu're absƖoυteεｌｙy 𝔯𝔤𝔤𝔥𝔱! 𝔜𝔲০'𝔢𝔯 𝔟ɑｓ૦𝔱ｕεⅼү 𝔯ι𝔤һｈ𝔱! Yo'urre 𝒷𝓈𝓁ℴ𝓊𝓉ℯ𝓎 riigh!t ｏu'ȩr ɑbb০̥͑ѕｌuteƖy rgιһt! у̷͌үս০'rr eab̸̀͜b͊͠slotｕεｙl іrgght̒͋!͔́̐ у҉͇͍u'reε ａbso͝εtllу іrg̐ḧ̕t̷ͥͫ!̶͑ ү̫̻̎ο̴͓͜ｕ̨̮'̮̆r̝̹͙ℯ аbｓ̶̿૦̜̲͝ｌͦ͡ս͜t͈еⅼу rͣ͜ıgｈ̯t͝! Y̖͞οͅս̞ͣ̅'̯͂͞ŗ͇ｅ̦ а͟b̤͚͔ѕ̽૦̟̇ｌ̠ｕ̧t҉eƖу̦̅ r͖ｉ̴̄g̿͜һ̶t͘͞!
One idea or rule of thumb is perhaps that we should or could interact with an LLM only in the same ways that we would interact with a non-coauthor colleague... You might ask your colleague to help speed up your code, but you wouldn’t ask your colleague to write your code.
I don't follow this bit though! Why wouldn't you ask your non-author colleague to write your code, especially if they were pretty good at it and you could verify that the result was correct?
David Hogg's article on LLMs in Astrophysics, "Why do we do Astrophysics?" is the most thoughtful critique I've yet read on the topic, and relevant for computer scientists and software developers: https://arxiv.org/html/2602.10181