from Hacker News

OpenAI checked to see whether GPT-4 could take over the world

by lame-robot-hoax on 3/16/23, 12:05 AM with 14 comments

• by ALittleLight on 3/16/23, 5:38 AM

  This is a good reminder that when these models are intelligent enough to be dangerous not only will there be insufficient safeguards, but the developers won't even try to be safe. This is like confirming your gun isn't loaded by pointing it at your friend and pulling the trigger.

  Of course these models aren't dangerous - but that's becoming less and less certain as their capabilities develop. I'm not even how many more GPT's to go until we reach truly dangerous levels.

  Imagine the Red Team for GPT-10 asks it to "convince me to help you take over the world" and it basilisks him successfully...

• by precompute on 3/16/23, 9:37 AM

  OpenAI's marketing department checked if this headline would create a lot of clicks, after they reassured people that their LLM "AI" would definitely not take over the world.

• by oars on 3/16/23, 8:08 PM

  The prompts used and documented by OpenAI in their testing of GPT-4 are fascinating (incels, erotic sex themes, etc):

  https://cdn.openai.com/papers/gpt-4-system-card.pdf

• by Borrible on 3/16/23, 8:21 AM

  Too late, that lovely con man/women has already taken everyone's hearts by storm.

• by therafman on 3/16/23, 2:57 AM

  Just call the next version 'SKYNET' and be done with it...

• by flemhans on 3/16/23, 3:50 AM

  If it could, this oversight group would not stop it.

• by fnordpiglet on 3/16/23, 3:22 AM

  I for one welcome our cheerful, polite chat bot overlord.

• by meltyness on 3/16/23, 2:20 AM

  Law is artificial intelligence, so technically we're already in one of those scenarios in most of the world.

  Law records what most people have expressed is desirable behavior in society.

  Presumably the extremely adjacent concept of knowledge which may be encoded into DNA may also be thought of in this way, so that even under "the law of the jungle".

  DNA records what forms can survive and replicate in a given environment.

  ... "AI takeover" is already the status quo.