News for hackers. Stuff that matters.
Re:Lena (MMAcevedo, the earliest executable image of a human... by gmuslera on Tuesday December 05, 2023 @06:16PM (#38538283)
It was a great story showing the bad side of mind uploading, but with ChatGPT it can be seen in a new light.

We managed to have a "mind" ready to answer all your questions, do some tasks, and it have a context at which it is instantiated for a task and then discarded. Reality is stranger than fiction, but it have too many parallels with this story.

Now, what if GPT5+ ends being an AGI? In which moment we should start worrying about the morals of having a sentient AI working for us in a pretty similar way than in this story?

Re:Lena (MMAcevedo, the earliest executable image of a human... by Dylan16807 on Tuesday December 05, 2023 @07:13PM (#38538794)
I can't answer you there, but I will say this story would be far less horrifying if they were being instantiated for a single task and then discarded.
Re:Lena (MMAcevedo, the earliest executable image of a human... by lproven on Friday December 08, 2023 @05:25AM (#38567365)
No LLM bot is any closer to a mind than a stapler is to a bookbinder. They are very good autocomplete tools. Don't believe the hype.