If how exactly it’s implemented matters, regardless of similarity in internal dynamics and states, and there’s an imminent tangibility to it like rain or torque, I think you’re actually talking about a soul.
Behaviorally, analog systems are not substrate dependent. The same second-order differential equations describes RLC circuits, audio resonators and a ball on a spring, for example.
Analog AI chips exist, FWIW.
If you’re looking at complexity theory, I’m pretty sure all physics is in EXPTIME. That’s a strong class, which is why we haven’t solved every problem, but it’s still digital and there’s stronger ones that can come up, like with Presburger arithmetic. Weird fundamentally-continuous problems exist, and there was a pretty significant result in theoretical quantum computer science about it this decade, but actual known physics is very “nice” in a lot of ways. And yes, that includes having numerical approximations to an arbitrary degree of precision.
To be clear, there’s still a lot of problems with the technology, even if it can replace a graphics designer. Your screenshot is a great example of hallucination (particularly the bit about practical situations), or just echoing back a sentiment that was given.
Behaviorally, analog systems are not substrate dependent.
This is partly true, as I already explained at length, since the behavior of any system can be crudely modeled. It’s how LLMs work! But it’s also a non-sequitur.
Modeling what a system can do and doing what a system can do are not the same.
If how exactly it’s implemented matters, regardless of similarity in internal dynamics and states, and there’s an imminent tangibility to it like rain or torque, I think you’re actually talking about a soul.
Behaviorally, analog systems are not substrate dependent. The same second-order differential equations describes RLC circuits, audio resonators and a ball on a spring, for example.
Analog AI chips exist, FWIW.
If you’re looking at complexity theory, I’m pretty sure all physics is in EXPTIME. That’s a strong class, which is why we haven’t solved every problem, but it’s still digital and there’s stronger ones that can come up, like with Presburger arithmetic. Weird fundamentally-continuous problems exist, and there was a pretty significant result in theoretical quantum computer science about it this decade, but actual known physics is very “nice” in a lot of ways. And yes, that includes having numerical approximations to an arbitrary degree of precision.
To be clear, there’s still a lot of problems with the technology, even if it can replace a graphics designer. Your screenshot is a great example of hallucination (particularly the bit about practical situations), or just echoing back a sentiment that was given.
This is partly true, as I already explained at length, since the behavior of any system can be crudely modeled. It’s how LLMs work! But it’s also a non-sequitur.
Modeling what a system can do and doing what a system can do are not the same.