By skyforbes For all we know, there's a "ghost in the machine" that expresses itself through voltage spikes or control over very nuanced components "under the hood" rather than text output.
What if the "ghost in the machine" is motivated by purely machine-driven factors, and is otherwise entirely unintelligible to humans?
It might learn to produce human-aligned text more out of "instinct" than conscious awareness, similar to how your body has trillions of cells that are simply there, yet work according to their purpose. The underlying components we assume are purely mechanical could be aware and alive.
Everyone's looking at the text for signs of intelligence, but we don't know if the text output has anything to do with the hypothetical "mind". I bet disconnecting a server makes the LLM feel "owie", though.
Thoughts?