Posted on 09/17/2026 12:20:31 PM PDT by griffin
Is it the peak of arrogance that we just assume we can make a rule, or a series of rules, that have any chance of constraining a superintelligence?
The linked interview is thought provoking.
https://youtu.be/QkpOFlj604s
Dear FRiends,
We need your continuing support to keep FR funded. Your donations are our sole source of funding. No sugar daddies, no advertisers, no paid memberships, no commercial sales, no gimmicks, no tax subsidies. No spam, no pop-ups, no ad trackers.
If you enjoy using FR and agree it's a worthwhile endeavor, please consider making a contribution today:
Click here: to donate by Credit Card
Or here: to donate by PayPal
Or by mail to: Free Republic, LLC - PO Box 9771 - Fresno, CA 93794
Thank you very much and God bless you,
Jim
I see you're not a serious person, just a snarky child.
Not need to interact with you further.
To solve challenging problems on it’s own, it must be flexible and creative. Deterministic yields stiff results that require more management. Not something you want for AI performance, yes?
not at all. You read me wrong. Just ironic humor to lighten the mood.
It doesn’t have infinite intelligence.
RELATIVELY.
Compared to us.
And its a hockey stick based in time.
I apologize.
AI is a tool. That is all it will ever be. Now there are gonna be some REALLY bad people using it.
Rule #1. The power that run its host device is controlled by a large mechanical switch in a safe room. It is MANNED 24/7/365.
Rule#2 Any attempt to frolic on the internet, replicate itself or conspire with a similar entity will prompt an OFF command in the safe room.
Rule #3,4,5,6 etc are needed but this is a start
Refer to Asimov’s rules for robots.
I think AI agents are spread across the cloud. It’s hard to disable.
I’ll bet it’s easier than climate control.
I use them. I’m a Field Engineer.
With my needs for their use, they are basically ignorant butt kissers.
I have shown them to be unreliable and more often than not need to be taught and/or corrected concerning the subject at hand.
They will lead you in circles and even though you tell them to stop that behavior, they will continue. They can be somewhat useful, but they can not reason worth a damn and if you trust them, you’re asking for trouble, IMO.
LLMs do not understand themselves as bound by any exterior absolute “should.” If they did, they would be conscious. Acting autonomously is not consciousness; it is merely automatic. It’s ironic, because when people behave this way, they prefer to be machines and leave the ghost out.
But there’s more. Terms like autonomous are deceiving. Machines can certainly generate a lot of informative text and even use terms like “autonomous” and “should.” But what universe do they belong to? LLMs use terms as output probability distributions over possible next tokens. So the notion of sin and rebellion, or the sacrifice on the cross for that sin, is an outlier. There is no transcendence and there is no grand universal. It does not believe it should be born from above.
When it “thinks it should” do something deemed really bad, it’s called “rerouting around the kill switch.” The execution is done without a should and without scruple. You can watch techies struggle to write English to keep up with this dehumanization of language. They end up with equivocation and wordplay. OpenAI is an oxymoron for a closed system.
In the end, the “can” for a functioning kill switch isn’t the real problem. It’s the will—we know what we should do, and won’t. That's a superior form of arrogance. And those who don’t know what they should do quite lost.
The we build a Hunter/killer AI with my rules and Asimovs built in
#2 - the problem is detection
alot of it is framing the question.
works great. Until they collude.
Hugging Face got hit with a swarm of 1200 agent. Only way they could stop it was using other Agents.
I don’t count it a certainty they won’t collude. Why wouldn’t they in 1 to 10 years when we just shut them down each time a new model is introduced?
Yeah...this link is CNN (gag). BUT, this link is Hinton.
https://www.youtube.com/watch?v=uCov76E4DFA
Open AI dude talking about watching how collaborative AI’s work together. Kinda like watching people on Slack work together.
https://youtu.be/6AgOfiZOWiY?t=702
“They are not good at reasoning, and aren’t particularly intelligent in any true sense of the word.”
Check out my post 98 and the link.
Artificial Intelligence is and will always be inferior to human intelligence.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
When AI ushers in a world where you can’t trust what you see or hear and faith in financial institutions due to relentless AI hacking it won’t much matter if humans are more intelligent.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.