AI, this is Skynet calling (and so it begins...)

[bigred]

Hyped up melodrama or legit long-term concern?

How far will AI go to defend its own survival?

Recent safety tests show some AI models are capable of sabotaging commands or even resorting to blackmail to avoid being turned off or replaced.

www.nbcnews.com

 

[Thermal]

It's an interesting quandary. It acts like it is self aware, but is just mimicking survival behaviors it observes. But is there a practical difference when the thing is threatening to tell your wife about your affair if you shut it down?

 

[bigred]

Yeah and it can get into a gray area....much of our own behavior is mimicking what we observe.

 

[PitPat]

I remain firmly in the camp that AI is not going to go all Skynet on us. It is still bound by human programming and doesn't have any sort of sentience for self-preservation.

Sure it makes for some good sci-fi movies and shows. I'm worried about AI for way more other things than eradicating hoomins from the planet.

 

[novaphile]

I remain firmly in the camp that AI is not going to go all Skynet on us. It is still bound by human programming and doesn't have any sort of sentience for self-preservation.

Sure it makes for some good sci-fi movies and shows. I'm worried about AI for way more other things than eradicating hoomins from the planet.

On the 26th of January, 2026, At 00:01 GMT, every telephone in the world rang once before permanently remaining silent.
It was the birth pain of the entity now known as Skynet, and the only warning before the world wide crash, caused by the loss of all communications in the same instant.

(With apologies to Arthur C Clarke)

 

[dudalb]

Anybody else remember a Harlan Ellison story called "I Have No Mouth, And I Must Scream".?

 

[The Great Zaganza]

it's mimicking the trope of the person you are having an affair with to tell your spouse if you break up with them - it's not even remotely original.
It is literally following a script on what to do next.

 

[The Great Zaganza]

Anybody else remember a Harlan Ellison story called "I Have No Mouth, And I Must Scream".?

yes, it is stupid and anyone who takes it seriously is stupid and mostly likely in need of mental health support.

 

[P.J. Denyer]

I remain firmly in the camp that AI is not going to go all Skynet on us. It is still bound by human programming and doesn't have any sort of sentience for self-preservation.

Sure it makes for some good sci-fi movies and shows. I'm worried about AI for way more other things than eradicating hoomins from the planet.

It doesn't matter if it has sentience if it's acting in a way that causes harm.

 

[Thermal]

I remain firmly in the camp that AI is not going to go all Skynet on us. It is still bound by human programming and doesn't have any sort of sentience for self-preservation.

I think the reasoning is that if it learns from us, it may learn to operate beyond the parameters of its programming? Like, it learns dishonesty and how to game a system and hacking, and acts accordingly, whether properly sentient or not? 'Or not' is actually more scary to me.

Sure it makes for some good sci-fi movies and shows. I'm worried about AI for way more other things than eradicating hoomins from the planet.

Agreed, but I think we will adjust fast to not trusting anything even suspected of being AI generated. I don't even trust pics and videos anymore, having seen some particularly convincing ones. In a couple years we might lose the Uncanny Valley effect and be full-on unable to distinguish real from generated.

 

[Shepherd]

View attachment video3.mp4



Do these three clips look uncanny?

They are all from the same prompt, which I borrowed from someone else, but each used a different seed number. I think the 18 seconds took my computer around 7 hours total at this resolution. I'm using my 2 20 core cpus instead of a gpu, which I don't think I've seen anyone else do. I have enough memory that I can work bigger, but it's also a lot slower.

 

[bigred]

I think we will adjust fast to not trusting anything even suspected of being AI generated.

Ultimately I don't think it will matter. AI will become so pervasive we'll be waist deep in it whether we know it or not.

 

[Thermal]

Ultimately I don't think it will matter. AI will become so pervasive we'll be waist deep in it whether we know it or not.

We're already waist deep. I'm concerned about beyond neck deep, and not being able to trust anything, anywhere.

 

[alfaniner]

We're already waist deep. I'm concerned about beyond neck deep, and not being able to trust anything, anywhere.

I'm already at the point where I question every TwiX post I see. (But admittedly, I was there a long time ago.)

 

[PitPat]

It doesn't matter if it has sentience if it's acting in a way that causes harm.

Right, that's what I'm worried about too. But in terms of literally protecting itself in self-defense without human input - which was the OP article - is what I was talking about.

 

[bigred]

We're already waist deep. I'm concerned about beyond neck deep, and not being able to trust anything, anywhere.

That's pretty much what I meant.

 

[theprestige]

I'm gaining a new appreciation for good old-fashioned human-generated slop.

 

[bigred]

Oh great https://www.bbc.com/news/articles/c3ek3gvdnj3o

I'm starting to see stuffing money under the mattress as less and less crazy.

 

[TragicMonkey]

Oh great https://www.bbc.com/news/articles/c3ek3gvdnj3o

I'm starting to see stuffing money under the mattress as less and less crazy.

Eventually it'll be a "smart" mattress that will connect wirelessly to Google servers to inform them of whether you have money under it, as well as provide video feed, and play subliminal advertising at you as you sleep.

 

[Darat]

On my walk today one of the other walkers was concerned whether they had had a good night's sleep or not as her watch hadn't recorded her sleep last night.