Conversation

Notices

1. Embed this notice
   Martin Escardo (martinescardo@mathstodon.xyz)'s status on Tuesday, 26-Nov-2024 07:20:13 JST Martin Escardo

   So I tried to be open-minded, as always, and had some goes at ChatGPT in the last few months.

   To find something I wanted to buy. To write a Haskell program. To write an Agda proof. To have my VPN ignore ssh connections in MacOS. To write a recommendation letter. To improve my prose. And much more.

   All I managed to achieve was to waste time.

   There wasn't a single thing that made me think "Oh, yeah, this is great! I am adopting it."

   But, who knows. Maybe there are some kinds of jobs that ChatGPT can do better than people. I am sorry for those people, if they exist.

   • Embed this notice
     julesh (julesh@mathstodon.xyz)'s status on Tuesday, 26-Nov-2024 07:20:11 JST julesh

     in reply to

     @MartinEscardo The one thing I've found it definitely useful for is grant writing. That's a perfect use case for a bullshit generation machine

   • Embed this notice
     Ross Duncan (rossquantum@mathstodon.xyz)'s status on Tuesday, 26-Nov-2024 17:44:27 JST Ross Duncan

     in reply to

     • julesh

     @julesh @MartinEscardo Did any of those grant applications get funded?

   • Embed this notice
     julesh (julesh@mathstodon.xyz)'s status on Tuesday, 26-Nov-2024 20:54:23 JST julesh

     in reply to

     • Ross Duncan

     @rossquantum @MartinEscardo This remains to be seen, I only started doing it recently. The last ~20 or so I wrote the honest way were not funded so it can't be any worse. To be clear, I'm not running entire applications off it, this is for specific questions that I consider too stupid to be worth spending any brain on

   • Embed this notice
     Ross Duncan (rossquantum@mathstodon.xyz)'s status on Tuesday, 26-Nov-2024 22:12:26 JST Ross Duncan

     in reply to

     • julesh

     @julesh @MartinEscardo Maybe just automate the whole thing and spend your time doing research instead of writing grant applications. (Assuming some external force compels you to submit them in the first place). #donthatetheplayerhatethegame #internalcontradictionsofcapitalaccumaultion

   • Embed this notice
     julesh (julesh@mathstodon.xyz)'s status on Tuesday, 26-Nov-2024 22:30:07 JST julesh

     in reply to

     • Björn Gohla
     • Ross Duncan

     @6d03 @rossquantum @MartinEscardo We discussed the idea unironically. Nothing commerically available can do it for sure, but the general idea is "write enough by hand and then fine tune". Maybe it works, maybe it doesn't

   • Embed this notice
     Björn Gohla (6d03@mathstodon.xyz)'s status on Tuesday, 26-Nov-2024 22:30:09 JST Björn Gohla

     in reply to

     • julesh
     • Ross Duncan

     @rossquantum @julesh @MartinEscardo Building a machine, that writes winning grant applications could be it's own research project 😇

   • Embed this notice
     julesh (julesh@mathstodon.xyz)'s status on Tuesday, 26-Nov-2024 23:29:43 JST julesh

     in reply to

     • Björn Gohla
     • Ross Duncan

     @rossquantum @6d03 @MartinEscardo Oh, yeah. I'm doing this for Cybercat Institute, who actually need funding. If I was doing it for the university then yeah

   • Embed this notice
     Ross Duncan (rossquantum@mathstodon.xyz)'s status on Tuesday, 26-Nov-2024 23:29:44 JST Ross Duncan

     in reply to

     • Björn Gohla
     • julesh

     @6d03 @julesh @MartinEscardo The aim is not win the grant. The aim is satisfy your dean that you tried to get one. A much lower bar.