Human sociability evolved because the less sociable individuals were either abandoned by the tribe and died.

Once we let these models interact with each other and humans in large online multiplayer sandbox worlds (it can be text-only for all I care, where death simply means exclusion), maybe we'll see a refinement of their sociability.

It can carry a thread of conversation, no?

So once ChatGPT has the ability to contain state, it will be conscious?

This a common fallacy deriving from having low level knowledge of a system without sufficient holistic knowledge. Being "inside" the system gives people far too much confidence that they know exactly what's going on. Searle's Chinese room and Leibniz's mill thought experiments are past examples of this. Citing the source code for chatGPT is just a modern iteration. The source code can no more tell us chatGPT isn't conscious than our DNA tells us we're not conscious.

The "source code" is a 175 billion parameter model. We have next to no idea what that model is doing internally.

We have our own source code, but for whatever reason we still insist we are sentient.

If we are allowed to give in to such delusions, why should ChatGPT not be allowed the same?