I don't think the chat logs show anything of the sort. It still, at it's core, is just predicting the next word in a sentance. The worrying thing, as we move past the uncanny valley of AI chat bots who are close to human level but still clearly machines to AI chat bots who feel and act human, as that people will start to find it hard to tell the difference.
The Replika AI app is an example of this. It can have conversations which feel real, although it's still clearly a Chatbot, and people are already saying they're building a real connection with it. It currently uses an open-source AI tool which is very far of the cutting edge tools that are currently available, never mind the ones that haven't been publicly released yet.
GPT3 is already able to write convincingly about any subject, even ones which don't exist in real life! And, it can already write new programs, I've used GPT3 Codex to write computer programs in Java before and it's surprisingly good. Apparently Googles version is even better.
You do wonder how manipulative an AI Chatbot could be that acts human and is given the express goal of convincing as many people as possible of all sorts of different things. If AI can already convince people that they have sentience over a series of chat logs, what could they do in the future when the technology becomes even more advanced.
It's a pretty scary thought.