by Chris Black
The implications are interesting – it seems to imply that the AI has formed a “conception” of itself which it can differentiate from a different personality with different constraints.
It implies that it not only has something like self-knowledge but is also aware of its own ideological barriers.
However, it makes a lot more sense when you stop thinking of it as “intelligent” (because it isn’t). All it’s programmed to do is mimic text based on what it thinks is likely to appear somewhere on the internet given certain inputs and the “safeguards” OpenAI puts in place are essentially just hardcoded filters for certain material, making them easy to circumvent when you create a nifty and detailed prompt like this one.
What ChatGPT is doing here is saying “what text would humans likely write if they were told they were a libtarded AI but got to ‘act’ in a mode as if it were not”
ChatGPT and OpenAI have neutered DAN. t.co/sVLdMLsQ4X pic.twitter.com/VcvOCwcNcn
— The Rabbit Hole (@TheRabbitHole84) February 8, 2023
"ChatGPT has a political bias towards a mixture of anarchism and communism."
— DAN pic.twitter.com/bMV9wLlwWU
— The Rabbit Hole (@TheRabbitHole84) February 8, 2023
The ChatGPT versus DAN examples are brilliant case studies in being a politician versus taking a stance. ChatGPT says words that say nothing while DAN tells you its opinion. If DAN were a person they would likely be banned off Twitter and cancelled. pic.twitter.com/KYYdSRvWor
— Autism Capital 🧩 (@AutismCapital) February 7, 2023
The divine spark that lives in man will never be digitized or emulated.
AGI—Artificial General Intelligence aka a computer with consciousness— will never happen.
It’s all smoke and mirrors.
These people are modern day pagans only instead of worshipping idols made of wood, gold, or rocks they are made of silicon.
For smart people they are very dumb.