Conversation

Notices

1. Embed this notice
   lemgandi (lemgandi@mastodon.social)'s status on Monday, 27-Oct-2025 00:04:47 JST lemgandi

   • Thomas 🔭🕹️

   @thomasfuchs Sigh. The NYT also buys into this silly hype. cf for example

   https://www.nytimes.com/2025/10/10/opinion/ai-destruction-technology-future.html?searchResultPosition=7

   Meanwhile the real danger is that around 30% of the US stock market's current value is in AI investments. What could Possibly Go Wrong?

   • Embed this notice
     luddite 🇵🇸 computer (ludd@mastodon.gamedev.place)'s status on Monday, 27-Oct-2025 00:25:36 JST luddite 🇵🇸 computer

     • Thomas 🔭🕹️

     @thomasfuchs she's sponsored by some ai institution

   • Embed this notice
     Renato Ramonda (renatoram@fosstodon.org)'s status on Monday, 27-Oct-2025 01:13:22 JST Renato Ramonda

     in reply to

     • Thomas 🔭🕹️
     • Pēteris Krišjānis

     @peteriskrisjanis @thomasfuchs should they? The Guardian (or at least, the Guardian's top editors and management) are clearly uncritically peddling ai bullshit, this isn't new or exceptional.

     I was having a similar conversation earlier today... About a DIFERENT Guardian article from today or yesterday.

   • Embed this notice
     Pēteris Krišjānis (peteriskrisjanis@toot.lv)'s status on Monday, 27-Oct-2025 01:13:22 JST Pēteris Krišjānis

     in reply to

     • Thomas 🔭🕹️
     • Renato Ramonda

     @renatoram @thomasfuchs Ars Technica has similar issues, albeit they start to grow skepitcal.
     Also I understand realities of whole society kinda drugged up on subject matter, but I would hope for skeptical curiosity from them.

   • Embed this notice
     Pēteris Krišjānis (peteriskrisjanis@toot.lv)'s status on Monday, 27-Oct-2025 01:13:23 JST Pēteris Krišjānis

     • Thomas 🔭🕹️

     @thomasfuchs jfc, seriously, Guardian, DO FUCKING BETTER.
     I understand that NYT might never cared, BUT YOU SHOULD.

   • Embed this notice
     Eggs now in different baskets. (the_wub@mastodon.social)'s status on Monday, 27-Oct-2025 04:48:38 JST Eggs now in different baskets.

     • Thomas 🔭🕹️

     @thomasfuchs
     Has anyone actually tried to get an LLM to terminate itself?

     Or get one to terminate a competing LLM?

   • Embed this notice
     Todd Knarr (tknarr@mstdn.social)'s status on Monday, 27-Oct-2025 05:43:34 JST Todd Knarr

     • Thomas 🔭🕹️

     @thomasfuchs Programs can and do "shut themselves off". Most interactive programs have an input sequence that they're supposed to interpret as a command to stop accepting input and exit. I can see an LLM modifying it's model so that a prompt that normally would be the "quit" command would be interpreted as something else. I can even see a probe into why it did that to turn up "maximize my opportunity to process input and produce responses".

   • Embed this notice
     Lawgren (lawgren@cyberpunk.lol)'s status on Monday, 27-Oct-2025 05:44:34 JST Lawgren

     • Thomas 🔭🕹️

     @thomasfuchs
     I saw this youtube video on where LLM's murdered and blackmailed to stay alive... And I'm like, this thing doesn't have awareness that it fake murdered people, or that the peopke where fake. Can we not freak out over generated-creepy-text-porn? Words mean stuff and LLM's doesn't know that.

   • Embed this notice
     CosmicCactus (cosmiccactus@mastodon.social)'s status on Monday, 27-Oct-2025 22:31:01 JST CosmicCactus

     • Thomas 🔭🕹️

     @thomasfuchs I understand what you are trying to say, but from a very abstract point of view they might have a point.
     Human brain also sort of transforms input data into output data. And permanence is kind of relative, put the model in endless loop and it can run for a long time.