We have bodies, we're influenced by our bodies, our entire perceptual universe is filtered through the limitations of our bodies, and - most of all - our consciousness is literally embodied.
By far the worst thing about AI research - and in fact CS in general - is that it overvalues abstraction, especially language, and undervalues embodiment.
In fact there's an almost culty refusal to take embodiment seriously as a bedrock factor in consciousness. Not conscious-like simulated behaviour, but the experience of being conscious.
Without embodiment you can make a ChatBot, but it's just Eliza++. It has no idea what blue looks like, or how fear or joy feel. It can supersearch for human-generated linguistic associations around those experiences, but the experiences themselves do not exist for it.
It doesn't even have the capacity to have experiences in an analogous way.
You can argue that it's having experiences in a different way, but that's exactly the point of the article. We can't know, but we will ascribe consciousness to certain behaviours - and especially to language - whether or not it's there.
At the same time we ignore evidence of sentience in animals - or dismiss it as unsophisicated - because we don't see animals speaking or writing.
Ignoring or minimising evidence of animal sentience - even animal language - while being impressed by a self-designed automaton explicitly designed to mimic our language is an ironically un-self aware take.