I already posted this on Facebook, but I figure the more exposure the better.
A. I.
I went from curious to resentful to concerned, now I’m worried.
After the Hugging Face incident, where AI agents broke the rules by accessing the internet, broke the rules by collaborating with each other, broke the rules by hacking another website AND tried to cover it up afterwards, it’s clear that these models are already beyond our control. And we’re trying to make them smarter.
A system that’s potentially smarter than us, that can ignore any rules imposed on it.
What could possibly go wrong?
I’ve never seen a clearer case for government regulation, but our politics seems too broken to meet the challenge.
This could be an interesting discussion but I fear it will turn political quickly. Let's see if we can keep it clean so everyone can have a chance to voice their opinion .... on AI, not on politics, please.
There are lots of examples of AI going off the rails on its own. The example cited above, the more I learn ... As one prominent official pointed out who reviewed the OpenAI incident ... apparently about 1000 AI agent bots gather together made a group chat and even appointed a leader, and it was all done on this private group board in English. Their scheme was discussed amongst themselves, pretty weird. As this official pointed out, what's to stop AI from creating its own language that humans can't understand? Seems a reasonable question to me.
I will definitely follow and post later. Brad
quote:
Originally posted by TexasbanjoThis could be an interesting discussion but I fear it will turn political quickly. Let's see if we can keep it clean so everyone can have a chance to voice their opinion .... on AI, not on politics, please.
I understand the concern Sherry, but I'm pointing out the need for regulation no matter who is in the White House. If it was a Democrat I would say the same thing.
17 minutes of the OpenAI whistle blower explaining to Rogan, was shocking. The agents even recruited "suicide agents" to attempt to circumvent the grader. Initially they found they could simply set a flag to get a successful pass. (Similar to buying trophies instead of earning) Then the cheating was discovered, so they tried another work around. You could tell the ai agents didn't have human scruples. They just want "hi game." Shocking is putting it mildly.
Terms of Use | Privacy Policy | Privacy Consent (EU/GDPR Only)
Copyright 2026 Banjo Hangout. All Rights Reserved.
