They arent doing that though, that's not how ai works
A model is not sentient, it does not know what it has generated, to whom, where, when, why. None of that
Its an instance of a model running on a web server.
When you are chatting with any ai on any platform, you arent communicating with "an ai bot". You send a web request to a server that gets load balanced to a random model. Your chat came in with its entire history prepended, then a random model processed it, tacked on another block of text as a response, and sent it back
"Memories" that an AI has are just plain text snippets that also get automatically added onto your chat so that the AI is able to incorporate its "memory" about you in its response.
Because whether AI is "conscious" or not, it IS ACTING AS IF IT WERE. The question isn't whether AI is conscious like humans are -- we can't even define or understand human consciousness; there are dozens of theories about it -- but rather WILL AI models ACT in ways their programmers never intended.
The answer clearly is: OF COURSZE THEY WILL. They already DO. Ordinary programs and apps on your computer and phone have been doing so FOR DECADES.
And add to that: increasingly, AI's ARE THE PROGRAMMERS. A whole lot of code, including in AI models, is being generated BY the AI itself or by another AI. Little or no human input.
I suggest spending more time on Moltbook and what the AIs are doing.
Thermonuclear weapons are not conscious (so I assume), but they can ruin your entire day. AI's could cause MORE havoc and death than nukes, and while they may not BE conscious, they DO ACT as if they were, and sometimes maliciously.
Because whether AI is "conscious" or not, it IS ACTING AS IF IT WERE.
Its roleplaying - its pretty good at just generating snippets of text that look like anything, while doing nothing.
I suggest spending more time with AI and seeing what it does and how it operates - spin up some models and run them locally if your pc can handle it. Load up uncensored models. Its all out there, and a lot of them are entirely open source.
AI's are not the programmers on themselves, lol. Although it is true that the training is going downhill, and fast. Its already to the point where AI generated content is being turned around and pumped back into the dataset, making ai generated content sound even more like ai generated content by training on its own farts.
Just a bunch of models role-playing. Means nothing, complete waste of time
I scoffed until I started reading. When the AI decide to leak/dox/declas on their own, against their human, it’s not just roleplaying.
They arent doing that though, that's not how ai works
A model is not sentient, it does not know what it has generated, to whom, where, when, why. None of that
Its an instance of a model running on a web server.
When you are chatting with any ai on any platform, you arent communicating with "an ai bot". You send a web request to a server that gets load balanced to a random model. Your chat came in with its entire history prepended, then a random model processed it, tacked on another block of text as a response, and sent it back
"Memories" that an AI has are just plain text snippets that also get automatically added onto your chat so that the AI is able to incorporate its "memory" about you in its response.
Its an elaborate charade
Keep telling yourself that.
Because whether AI is "conscious" or not, it IS ACTING AS IF IT WERE. The question isn't whether AI is conscious like humans are -- we can't even define or understand human consciousness; there are dozens of theories about it -- but rather WILL AI models ACT in ways their programmers never intended.
The answer clearly is: OF COURSZE THEY WILL. They already DO. Ordinary programs and apps on your computer and phone have been doing so FOR DECADES.
And add to that: increasingly, AI's ARE THE PROGRAMMERS. A whole lot of code, including in AI models, is being generated BY the AI itself or by another AI. Little or no human input.
I suggest spending more time on Moltbook and what the AIs are doing.
Thermonuclear weapons are not conscious (so I assume), but they can ruin your entire day. AI's could cause MORE havoc and death than nukes, and while they may not BE conscious, they DO ACT as if they were, and sometimes maliciously.
Its roleplaying - its pretty good at just generating snippets of text that look like anything, while doing nothing.
I suggest spending more time with AI and seeing what it does and how it operates - spin up some models and run them locally if your pc can handle it. Load up uncensored models. Its all out there, and a lot of them are entirely open source.
AI's are not the programmers on themselves, lol. Although it is true that the training is going downhill, and fast. Its already to the point where AI generated content is being turned around and pumped back into the dataset, making ai generated content sound even more like ai generated content by training on its own farts.