I suppose it depends on what you mean by "AI", and, IMO, would be very much dependent on the hardware/architecture of the thing. Is a long-running LLM all that

I suppose it depends on what you mean by "AI", and, IMO, would be very much dependent on the hardware/architecture of the thing. Is a long-running LLM all that different from your own neural-net? (Should we develop AI capable of training itself, granting it agency--or, leave well enough alone and emphasize training/building models which behave more like "smart glass"?) Humans and telescopes both have lenses... but, a telescope, ostensibly, cannot witness itself.