It's not about what you point, is about how you define "dog" (and it can get much worse than dog, e.g. freedom, humanity, etc).

It's amazing how people think they can understand/debunk/solve philosophical problems of ages with a moment's hasty thought that doesn't even understand the full issue in the first place...

... until a later bystander asks you, "What is a dog?" and you now have to define it without the associated dog present.

Or someone looks at another breed of dog and asks "What is that?" because he isn't aware of canine taxonomy and thinks only that original breed is a creature called "dog".

Welcome to Derrida's version of the halting problem. These are actual problems of NLP today.

That's not what people do in natural language processing. Don't appropriate the terminology of other fields for your own purposes and claim you're doing that field.

That's not what is done in NLP because that cannot be done by NLP... for the reasons that Derrida pointed out decades ago.

Don't tell NLPists what not to do, thanks.

In NLP, people don't just throw up their hands and say "grounding is hard, it's so hard it's impossible, let's not do it". Grounding is sometimes hard, but it's often perfectly tractable for a given problem domain.

If we haven't solved strong AI yet, don't count that as a victory for your unrelated field.

Faith in an AI clasification that is invented from a human mind bound by the biases Derrida described long before people could put faith in strong AI?

Sorry, I don't do faith. I do correlation. You just hate Derrida and can't apply any of his observations.

The human does not discover AI. It invents it. And all of the cognitive biases we have are passively passed on with every "axiom" of AI we create. The idea of even classifying AI is ontologically invalid... for the reasons Derrida already pointed out.