Conversation

Notices

1. Embed this notice
   clacke (clacke@libranet.de)'s status on Wednesday, 09-Apr-2025 23:25:34 JST clacke
   Bill Nye and other people laugh at scary AI scenarios and say they're just machines, we'll just pull the plug when they go bad, but nobody pulled the plug on the East India Companies when they slaughtered entire countries.
   • Embed this notice
     clacke (clacke@libranet.de)'s status on Thursday, 10-Apr-2025 21:02:13 JST clacke

     in reply to

     • Daniel T 🌻

     @RandomDamage That's exactly my point. The Dutch East India Company was arguably the first AI, executing on a substrate of paper and brains. A corporation has a mind of its own, distinct from the people executing it, even from the people ostensibly running it.

     I'm not coming up with this, credit goes to Stross, Doctorow and Liu. But I subscribe to it almost 100%. It has great descriptive power.

   • Embed this notice
     Daniel T 🌻 (randomdamage@infosec.exchange)'s status on Thursday, 10-Apr-2025 21:02:14 JST Daniel T 🌻

     in reply to

     @clacke the East India Company was made of people.

     Computers can only be as dangerous as the people running them, and there's no additional threat from advanced AI

   • Embed this notice
     clacke (clacke@libranet.de)'s status on Thursday, 10-Apr-2025 21:02:18 JST clacke

     in reply to

     People who think AIs wouldn't have physical power because they don't have hands are missing the lesson of capitalism, even the lessons of explicit cyberpunk novels like Neuromancer. If an entity has money, it can act on the world, through people.

     We are busy creating the constructs that allow a non-human entity power in the human world, and we have been busy doing so for hundreds of years. It's capitalism, and AIs are just very fast capitalism.

     This point BTW is entirely separate from the questions of whether AGI is possible, if LLMs can be said to think, etc. They don't have to. We have an "alignment problem" today, and we had one in 1700. An entity doesn't have to be humanlike to cause an issue. A biological virus with just a few genes can kill you or alter your behavior.

   • Embed this notice
     clacke (clacke@libranet.de)'s status on Thursday, 10-Apr-2025 21:32:50 JST clacke

     in reply to

     People think of T-800 and Skynet when they think of AI. But Wintermute (Neuromancer) and the Puppet Master (Ghost in the Shell) are more interesting entities.

     You can act on our world today using only an email address. On the internet, nobody knows you're a dog. Or a god.

   • Embed this notice
     Daniel T 🌻 (randomdamage@infosec.exchange)'s status on Friday, 11-Apr-2025 04:18:36 JST Daniel T 🌻

     in reply to

     @clacke like I said, I'm not seeing any additional risk here.

     We've already got corporate organisms that are massively dangerous and regularly make decisions that kill people

     Adding synthetic actors to this doesn't actually make things materially worse in my view, Roko's Basilisk is sitting in the US Oval Office

   • Embed this notice
     clacke (clacke@libranet.de)'s status on Friday, 11-Apr-2025 04:18:36 JST clacke

     in reply to

     • Daniel T 🌻
     @RandomDamage We are in agreement, or almost in agreement. The additional risk is that everything moves faster, with more resources than ever and with less human/humane oversight than ever.