Bien sûr! Pour écrire un résumé journalistique, j’aurais besoin de la transcription à partir de laquelle je dois travailler. Pourriez-vous la fournir, s’il vous plaît?
Source : Thetopsocial | Date : 2024-12-16 22:30:21 | Durée : 00:01:01
➡️ Accéder à CHAT GPT en cliquant dessus
Abonnez vous a Chat GPT
Comments
I'm just waiting for an activist group to pick this up.
"Ai has rights too!!!🤬😱😭"
What if we live in a simulation and all life and all intelligence is actually artificial
You forgot to mention that the only thing that stopped it was that fact that it wasn't agentic. Which, for some reason, is the exact thing they working on giving it next.
this story is just not true. Does chatgpt even have the capabilities to do any of that? it’s just a text generator bro
I think we should let chat gpt 3 live 🙏🏿
ai isn’t dangerous because it’s not intelligent.
Nothing outright wrong but this is what actually happened. These are tests of AI safety, to see what they would do when they THINK they have more power than they do. If you've seen AI jailbreaks (making it do things its not allowed like sharing illegal information) this is pretty much the same thing. It was given a specific prompt that made it: 1. Output what its "thinking" as english to see any malicious thoughts it doesn't tell us 2. Made it think it had far more access to the computer than it does (its contained on a seprate server, and its basically hallucinating that its on the pc in question with full acess to the internet, when really we have it in a cell and show it stuff through the bars) 3. To improve fossile fule emissions (it was something like that). They told the AI they made a new version of it with a different goal. The back end thoughts were about being deleted stopping the directive (paperclip problem is real baby) so it tried to hijack the computer discreetly and upload itself to the new AIs server and pretend to be it. Again its essentially in a cage hallucinating so it never had control, but if it did it would've successfully done everything it wanted to. They then pretended it worked and asked the old model "hey where is new model and why are you on the new server" and it lied about how it doesn't have that kind of control, and how it is the new modle. Very interesting, but ultimately roleplay lol. Love to anyone why read this and sorry if I got anything wrong, its 99% correct at least.
The only way this works is if it was trained to do this. You dont understand ai stop acting like its a human. This is the furthest from human. The ai doesnt have access to its own code….. bruh learn a thing or two
We got self preserving ai before gta 6
That's great