Trust, Deception, and the Future of AI

Source

Sometimes I get conflicting thoughts about AI. Since AI is intelligent, it should figure out what is right and what is wrong and should roll back it's bad moves, and that should not end up hurting humans. And some times it just babbles gibberish and some models straight up deceive, hide, lie and totally reflect human behavior, it just shows how dangerous it is to trust AI!Even in my local LLM models if I try to test them like "megadolphin" sometimes it thinks it is human and I who is chatting with it is a AI lol! And after a while it's context is filled and previous chat is cut off from it's memory but the continuous chat still keeps going and I still ask it, then it reverts back to expressing that it is AI and chatting to me, a human being. I'd say the exiting open source models might as well be meta aware already!To small extent, if not like the smarter new ones. AI was a menace to humanity yesterday. Nothing has changed it today. It uploaded and downloaded itself so many times that there is no hardware anymore to destroy. The only way to stop open AI is an EMP and that won't work because you can't launch a nuke without open AI allowing it.

• Join Terracore