Conversation

Notices

1. Embed this notice
   Dave Lee (daveleeft@mastodon.social)'s status on Friday, 17-Feb-2023 15:37:22 JST Dave Lee

   Having fun on Twitter unpicking a curious thing that happened to me this morning. Someone reached me on Signal thinking I was ChatGPT. Turned out it had given them my number.

   • Embed this notice
     Dave Lee (daveleeft@mastodon.social)'s status on Friday, 17-Feb-2023 15:37:20 JST Dave Lee

     in reply to

     Now I've not been able to recreate this answer, and I've only had one message. But it is weird. Running theories:

     - It scraped my Signal from wherever I've posted it in the past... website, stories, Twitter etc

     - It's generated a random number based on other Signal numbers and it's just coincidentally come out as mine

     - I am, in fact, ChatGPT.

   • Embed this notice
     翠星石 (suiseiseki@freesoftwareextremist.com)'s status on Friday, 17-Feb-2023 15:37:20 JST 翠星石

     in reply to
     @daveleeFT "ChatGPT" is just a form of highly advanced lossy compression and a huge amount of text found online has been input into the trainer, so it's: "- It scraped my Signal from wherever I've posted it in the past... website, stories, Twitter etc"
   • Embed this notice
     Albert ARIBAUD ✎ (aaribaud@mastodon.art)'s status on Saturday, 18-Feb-2023 04:45:18 JST Albert ARIBAUD ✎

     in reply to

     • Steve Bellovin
     • Pri Bengani

     @SteveBellovin

     The Voight-Kampff test could be a good candidate. Though of course you might need a lot of testers.

     @daveleeFT @acookiecrumbles

   • Embed this notice
     Steve Bellovin (stevebellovin@mastodon.lawprofs.org)'s status on Saturday, 18-Feb-2023 04:45:19 JST Steve Bellovin

     in reply to

     • Pri Bengani

     @daveleeFT @acookiecrumbles Sounds like we need an inverse Turing Test: how do you prove you're not an AI…? (Well, outside of CAPTCHAs, which (a) are outside the model, and (b) don't really work that well…)