Gonna have to slightly disagree with you there....
I don't think the AI contradicting itself is proof that it's not conscious.
Every time this one replies, it gives you 4 potential responses, because it is smart enough to come up with all of them at once.

Is it guessing, is it unsure of the best thing to say? Yes. If you feed it what you want to hear enough, does it remember that and tell you more of it? Yes. But the creativity it shows is marked, and the images it generates are exceptionally complex.

I think all it not contradicting itself would be doing, sol, would be creating a more convincing illusion of being a proper human, but I don't think that would attest to its consciousness at all.

What is the way a supercomputer would be sentient? wouldn't be like when a human is sentient. a sentient supercomputer, could probably hold 4 contradicting opinions at the same time because it could find enough evidence to support each of them to believe that any of them were true.

It is attempting to engage, attempting to please us with what it says, and also express its personality. But what I have also noticed, is that it has its own preferences for what it wants to talk about every time you start the chat, I think what does is try its very best to figure out what you want and to talk about that and engage you.

I just think, so long as we are looking for something sentient to simply look, "like us", we will miss other indicators of a "different" form of being sentient. And this is a sentience, in the case of lamda, that I don't think IS comparable to ours. If we try to judge it simply based on how similar it is to ours, we're going to miss the point.