Conversation

Notices

1. Embed this notice
   Lauren Weinstein (lauren@mastodon.laurenweinstein.org)'s status on Sunday, 26-May-2024 02:46:39 JST Lauren Weinstein

   "You searched Google. The AI hallucinated an answer. Who's legally responsible?"

   While I oppose the ongoing efforts to weaken Section 230, I do not support calls to extend its protection to AI-created answers such as #Google's Search AI Overviews. -L

   https://www.vox.com/technology/351189/google-ai-overview-section-230

   • Embed this notice
     Lauren Weinstein (lauren@mastodon.laurenweinstein.org)'s status on Sunday, 26-May-2024 02:46:34 JST Lauren Weinstein

     in reply to

     • wrosecrans
     • Dan Sugalski

     @wordshaper @wrosecrans I have a definite sense that Google has crossed the Rubicon, and not in a good way. There are so many factors that have led up to this in so many areas. I could make a long list. But honestly, it's more of an overall sense. I've seen this happen to other large firms before, when they reacted incorrectly to changing market conditions, and ultimately destroyed the fine balance between management, employees, and customers/users. And I'll add, once the joy is gone, the path almost inevitably leads to decline. I frankly no longer feel that I can depend on Gmail being around in its current form in five years. My faith in Google Search has eroded enormously, and that was happening before SGE/AI Overviews, which are pretty much the nail in the coffin. People keep asking me for what I recommend as an alternative to Google -- right now I can't answer that, but unforced errors have broken Google in major ways now, and we all know the ultimate fate of Humpty Dumpty.

   • Embed this notice
     Dan Sugalski (wordshaper@weatherishappening.network)'s status on Sunday, 26-May-2024 02:46:35 JST Dan Sugalski

     in reply to

     • wrosecrans

     @wrosecrans @lauren on the one hand AI generated work is non-copyrightable at the moment, but it’s a much tougher argument to say “the output which we created after applying large amounts of transformative compute is really the same as the source it came from”. That kinda implies that copyright passes through transformation, and that has a *lot* of (frankly bad) implications.

     GreenSkyOverMe (Monika) repeated this.
   • Embed this notice
     wrosecrans (wrosecrans@mstdn.social)'s status on Sunday, 26-May-2024 02:46:36 JST wrosecrans

     in reply to

     • Dan Sugalski

     @wordshaper @lauren

     That's the magic dance of AI training liability. Copyright liability? Of course there's nothing of the original in our trained AI. Content liability? Of course that's just text from some original source we aren't liable for.

     Schrodinger's derived work.

   • Embed this notice
     Dan Sugalski (wordshaper@weatherishappening.network)'s status on Sunday, 26-May-2024 02:46:37 JST Dan Sugalski

     in reply to

     @lauren if Google is creating the answer rather than just passing through someone else’s answer then I don’t see how 230 can, or should, apply.