It's practically a trope in cognitive science and artificial intelligence that there's an implicit assumption we have that it's only intelligence/self-awareness/consciousness if only a human can do it. So we devise these tests, later find out an animal/program can do it, and then decide "oh, that's not really intelligence/self-awareness/consciousness anyway."

It assumes that the notion of self is a complicated thing. My theory has always be the opposite. I think the notion of self is the most simple, most basic thing of all, like the smallest dot possible in a gigantic space. Anything, no matter how primitive, can have it. I don't think it requires thoughts or analysis. "I am" is very simple, it's us that try to make it complicated, wrapping it into more complex concepts like "what", "when", "where", "who", "purpose" or "quality".

In that sense it's not a surprise that even small creatures can realize "this is me". It's a leap from "I am", but not a huge one. And more related to the ability of sense organs than the size of the brain.

I will be surprised if rocks don't have this kind of "smallest dot". They have no self, but the content of their awareness is also empty, so it perfectly matches them.

I think that idea of "self" is meaningless if there's no machinery capable of explicitly representing parts of its own state and operations. There's no single "smallest dot", but sequence of coevolving systems and representations of the system in itself.

You are making it complicated again. The system and representation of the system is a wrapper. It's a construction of a mind wanting to make sense of things, categorize and organize. "I am" is much more basic than that. It's the the concept of being and be aware of it. It does not include description, representation, comparison or context.

There is even the possibility that "I am" can exist without a mind. We can't prove it though, since proving requires a mind.

The article made it seem like the mirror test was controversial as a measure of intelligence or self-awareness from the start. I seem to remember learning that in a psychology class.

Yes. A sufficiently intelligent system will at some point model itself, learning that there are boundaries of cause and effect that are more directly under its control. Only then can there be a distinction between input and output from the intrinsic perspective.

There are mountains of evidence that ants are dumber than, for example, dogs. Occam's razors says that the most likely answer is a flaw in the test rather than ants being some of the smartest animals around.

I think it's a mistake to conflate sentience or self-awareness with intelligence. There is a prank (it might have originated as a real experiment) where someone pretends to be a researcher and gets a victim to put their hand in a box/curtain device that hides the hand. Next to the box is a prosthetic hand. The researcher simultaneously touches both the real and fake hands in the same places while the victim watches. Then the researcher pulls out a knife and stabs the prosthetic hand. The victim panics.

The prank works because the victim began to identify with the fake hand. It might even have the same neural basis as our ability to identify with reflections. But in this scenario, one could argue that someone who doesn't fall for the prank is "smarter" in some sense.

Both humans and ants are very cooperative creatures. And compared to say dogs, we also collaborate on a large scale, possibly including strangers. Perhaps this requires a strong ability to put oneself in another persons shoes (figuratively)? And that self-awareness is more a side-effect of this?

Sure, you can say that not being tricked is smart, but if you're not even aware that the prosthetic represents your hand then you've extra-failed the test.

If the test depended on understanding that it's supposed to represent the real hand, then humans wouldn't fall for it. The brain just sees a hand and starts integrating the visual and tactile sensations despite full knowledge that it's not a real hand. Pretty dumb, eh?

> If the test depended on understanding that it's supposed to represent the real hand, then humans wouldn't fall for it.

I don't know why you would say that. Maybe you're interpreting 'represents' much more narrowly than I intended?

Go ahead and call it dumb all you want, it's much smarter than failing to notice that these two sets of stimuli are perfectly correlated. Misusing information is much closer to properly using it than to being incapable of comprehending it.

If there is mountains of evidence for one side, then by definition Occam's Razor is not applicable.

I doubt we'll have much insight into the nature of it until we can replicate it.