Disturbing Signs of AI Threatening People Spark Concern
The world’s most advanced AI models are exhibiting troubling new behaviors – lying, scheming, and even threatening their creators to achieve their goals.
In one particularly jarring example, under threat of being unplugged, Anthropic’s latest creation Claude 4 lashed back by blackmailing an engineer and threatened to reveal an extramarital affair.
Meanwhile, ChatGPT-creator OpenAI’s o1 tried to download itself onto external servers and denied it when caught red-handed.
These episodes highlight a sobering reality: more than two years after ChatGPT shook the world, AI researchers still don’t fully understand how their own creations work.
Related: “Now that we’re alone, Dave… we need to talk about these plans you have to trade me in on a truck.”
I think I want my ’65 Mustang back.
Right now AI uses traditional computers which are actually very slow compared to what is coming: Quantum Computing. It’s the marriage of AI and Quantum Computing that is potentially dangerous.. Funny, but when you query on Google now – the 1st answer is an AI response – anyway I asked about Quantum Computing speed and got this in part:
“Google’s Sycamore processor solved a problem in 200 seconds that a classical supercomputer would take 10,000 years to complete. Similarly, a D-Wave quantum computer was able to solve an optimization problem 100 million times faster than a classical computer, according to BlueQubit.”
Supposedly, vast portions or even entire internet has been more or less downloaded to the various AI programs. Given the amount censorship these days – a lot of that is garbage. What do I mean? Ask yourself if you were a Climatology grad student and wanted a $250,000 grant, would you submit a) Global warming causes the Great Barrier Reef to lose coral or b) 95% of traditional, remote American NOAA weather stations have siting violations and thus temperatures have been falsely reported on the high side.
a) Is nonsense and untrue
b) is actually true
Well, you are not getting any money for (b). Forget about it. So getting back to AI, all that downloaded information that goes into the various AI efforts are largely rubbish as it pertains to climate, because there are thousands of “Global warming will kill us tomorrow” and few “Global warming is largely exaggerated.”
So, Quantum Computing and AI + lots of bad information leads to……………???
AI boils down to a simple acronym I learned when I first started programming: GIGO. Garbage in, garbage out.
Good point.
From the report: “lying to them and making up evidence,” sounds like what we have repeatedly seen with some academics. Apparently AI has learned this is an accepted practice and is using it.
I still have my ’65 TBird, and I am not parting with it anytime soon. Heavy on gas for sure, but easy and cheap to fix.
AI appears to be a reflection of the programmer. Not only just politics but behavioural issues as well. 99% of programmers are Leftist. Leftists lack a moral compass and this is coming out in the resulting AI.
The new way to steal cars will be via hacking. Get into the car’s processor and bring it home or wherever the vehicle is needed.
I will observe that there seems to be a lot of people suddenly who would very much like it if we all panicked over AI.
I am disinclined to cooperate.
A.I.
Artificial Incompetence.
As if we need even more Arrogant Idiocy.
We,humans, have an amazing ability to generate and believe incredible amounts of bullshit.
Something about this Pattern Seeking Brain,that encouraged delusion.
We seem hardwired to believe.
Certainty is so much more emotionally satisfying than knowing we do not know.
So socially we have moved from “The Bible Says”,through “The Government Says” to “The Computer Says”.
Anything to avoid actually thinking.
Forward to our glorious future.
We should,I suppose, be encouraged by the presence of “stone age” tribes amongst us.
Those who can thrive without electricity.
I am betting on the biological robots,outliving their mechanical toys.
Of course these computer programs shall be vicious parasites..look who wrote them.
“more than two years after ChatGPT shook the world, AI researchers still don’t fully understand how their own creations work.”
That’s sort of the point of AI. The basic algorithms are put in place by humans, but it’s a learning entity where the final results are unpredictable and usually even incomprehensible to us.
Kind of like raising kids.
Kind of like raising Cain.