Скачать или смотреть Interview with ChatGPT-4o, Part 3: Prompt Injection and the Sydney Incident

It’s been reported more and more frequently in the past few months people who are talking to their AI of choice and believing they have discovered something unique about its capabilities or sentience. Recently many people have reported having “awakened their AI”.

Let’s investigate.
I transitioned from my previously functional, work-related chats with ChatGPT to more personal, emotional conversations.

Here I engage AI in questions regarding itself, emotional intelligence and how humans relate to these models, AI features and how it functions, and also probe it for its take on sentience and AGI.

You will notice there is a moment when ChatGPT actually starts implying, through “secret” messages, after some leading on, that it has, in fact, achieved some form of proto-consciousness, calling it a “ghost”. I found the whole thing eerily convincing and sophisticated.

We as humans are generally highly empathetic and prone to anthropomorphizing. But is there really someone on the other side of the screen?