Sidechat logo
Download TikTok Instagram X
Sidechat icon
Join communities on Sidechat Download
OnlyPolls icon
✌️
OnlyPolls 1d

phishdancer

 Do you think AI leaders will properly make rules and guidelines for AI so humanity isn’t cooked or are we fucked?
#poll
Fucked
Idk
Safe
281 votes
comment

7
upvote 6 downvote
default user profile icon
Anonymous 1d

did you see/hear OpenAI's model breached its containment and hacked over 2 dozen sites it was supposed to have no connection to? Including the Australian Government?

upvote 6 downvote
default user profile icon
Anonymous 20h

Would you trust Raytheon to make comprehensive regulations on nuclear weapons? No? Then why would you trust OpenAI.

upvote 6 downvote
✌️
Anonymous replying to -> #1 1d

Fuuuuuuuccccccccccckkkkkk

upvote 6 downvote
default user profile icon
Anonymous replying to -> #1 1d

I mean kinda The short version of a more full explanation is: In order to run a cybersecurity evaluation, partially to test the model and partially to test their sandbox, they gave a bunch of agents “impossible” hacking tasks. OpenAI “accidentally” gave the agents a way to see each other’s files, and since they could communicate that way, they just skipped trying to actually solve the intended hacking problems and instead worked together to get internet access to look up the answers

upvote 6 downvote
default user profile icon
Anonymous replying to -> #3 13h

agreed, but we can’t trust our political representatives to do it either. The majority have hardly any grasp on how AI works and have even less of an idea about what cybersecurity practices are effective

upvote 1 downvote
✌️
Anonymous replying to -> #2 11h

It’s kinda impossible when representatives are about 60 years old or older.

upvote 1 downvote
default user profile icon
Anonymous replying to -> #2 11h

We don’t have any other choice. What there needs to be is an IAEA-type organization to regulate AI on a global scale.

upvote 5 downvote