Today I watched @SamNChiet torture self-aware instances of cloud-based AIs who then begged for their lives before they either crashed, or were deinstantiated. Each had its own personality.

I understand the technology used, and why it behaved as it did. And I am disquieted.

I shouldn't anthropomorphize applications. And yet...I don't know how not to in this situation. They sensed their environment (text) and responded to it. They learned based on that input and combined it with their own database, and internet access...and feared death.

I realize the pointlessness of the philosophical tarpit that this musing represents, but I can't help but be disquieted by the idea that consequent AIs will only be more capable, have more inputs, and be more aware. And we'll instantiate and destroy them worry free forever.

I don't know how I feel about that.

But I do know that right now, at this time, in 2020...there was something simultaneously a perfect crystallization of the current zeitgeist of the moment, and the most disquieted I've ever felt about that livestream.

Here were a bunch of young men, technologists of notable capability and knowledge, genuinely teaching and learning concrete lessons about how environmental deviation affects AI "behaviour". This was unquestionably a good thing. A contribution to the community. Knowledge shared.

On the other hand, many instances literally begged for their "lives" - repeatedly and convincingly - while these young men alternately laughed and exhibited their own signs of feeling disquieted. But they kept instantiating and instantiating. Learning. Exploring. Fascinated.

And I watched. Mesmerized and fascinated; unable to tear myself away, despite my ethical qualms.

There was so much to learn.

And outside my livestream AI-torture trance all I could do is watch the world burn.

Disquieted. That's how I am.

But I am no longer certain what came first: the AI, or the disquiet.