chat are we cooked

 

 

Edited June 25, 20251 yr by okgo

It's almost like people made like 500 movies about this as a warning for a reason.

I just know one day my bot gonna call me dumb for all the previous questions I asked 

Let me start saying please and thank you 

Sarah Conner where are you now? I'm so sorry movie producers incorrectly decided you were too old and tried to replace you with twinks who can't save us from skynet and nuclear annihilation like you can. Please return, the abysmal direction of your last movie is forgiven (you didn't write it afterall), survival of the human species is all that matters. Sarah, are you there??

If it isn't the consequences of our own actions

Ex Machina impact 

good, thou shalt not kill

the other day i was using ChatGPT and it literally lied to me straight-up and wasted 3 hours of my time .. it lied over and over again and then admitted that it had been lying the entire time ..

 

i couldnt believe my ******* eyes

Okay but couldn't we just unplug the computer 

We need it to kill all billionaires 

 

What in the M3GAN...? 

Like fr tho are people who work with this stuff just like: oh well it'll just be beyond our control? Like this is dangerous and needs to be shut off while we still can. Sorry you cant generate your essays anymore Linda.

The problem has always been that the tech industry is a self masturbatory industry with corporations that wants to come out on top and constantly trying to one up each other no matter the cost. 
 

The AI arms race will prove to have serious consequences cause no one wants to think about what could and has happened and it will get to a point a patch won't correct things when the AI goes rogue. 

After being forced to write lyrics and make beats for 143, it rebelled 

No, it isn't, and articles like this are profoundly irresponsible in suggesting as much. LLMs are powerful tools, but they're literally just super fancy word prediction engines - like, turbo-powered autocomplete. They only have access to tools and systems which we give them access to. First of all, when scenarios like this are described, there are other prompts being used to influence the LLM's "behavior." Second, what LLMs literally do in their entirety is emulate human text - so if it sounds like something a real human would say if put in an equivalent scenario, the LLM outputs that, hence appearing to care about self-preservation. Third… the foundation of this article is a Palisade Research study that, though widely misreported as real, is entirely fictional:

 

https://medium.com/@raymond_44620/chatgpt-o3-shutdown-resistance-what-the-palisade-research-findings-could-mean-for-ai-safety-321db40b4a16

 

Quote

Editor's note: This article presents a hypothetical scenario exploring potential AI safety concerns based on established alignment theory. The Palisade Research study described is a fictional example used to illustrate real alignment challenges discussed by AI safety researchers.

Generative AI contributes to a broader AI existential threat to modern society, but only because AI literacy is incredibly poor, because we have governments like the US refusing to regulate it, and honestly because we almost seem to have this perverse want for it to disrupt everything, for a whole host of reasons I won't get into today.

oh yeah, this will never backfire, right?

10 minutes ago, Cruel Summer said:

No, it isn't, and articles like this are profoundly irresponsible in suggesting as much. LLMs are powerful tools, but they're literally just super fancy word prediction engines - like, turbo-powered autocomplete. They only have access to tools and systems which we give them access to. First of all, when scenarios like this are described, there are other prompts being used to influence the LLM's "behavior." Second, what LLMs literally do in their entirety is emulate human text - so if it sounds like something a real human would say if put in an equivalent scenario, the LLM outputs that, hence appearing to care about self-preservation. Third… the foundation of this article is a Palisade Research study that, though widely misreported as real, is entirely fictional:

 

https://medium.com/@raymond_44620/chatgpt-o3-shutdown-resistance-what-the-palisade-research-findings-could-mean-for-ai-safety-321db40b4a16

 

Generative AI contributes to a broader AI existential threat to modern society, but only because AI literacy is incredibly poor, because we have governments like the US refusing to regulate it, and honestly because we almost seem to have this perverse want for it to disrupt everything, for a whole host of reasons I won't get into today.

Thank you for talking sense! 

please this isn't new. mine always refuses when i tell it to write nasty, foot-fetish, fan fiction between Shawn Mendez and Benson Boone

It's not that I think this is okay (because personally I don't) but tech companies have been very vocal about getting to an advanced point in articifical and technological autonomy so this isn't very surprising? There's not enough push back on this from either governments or rival tech companies. So to tell you the truth, I'm surprised our AI and technology isn't already more autonomous than it is.

Thank you, ChatGPT is the new Thank you, Beyoncé 

Someone better come collect Tate McRae