No, hard disagree.
It is not a toy, because language is not a toy. Language is a ridiculously powerful tool we have as humans, we just take it for granted because we are very good at it.
What ChatGPT allows us is to evaluate the usefulness of language without all the other intellectual tools we have, like mathematics, logic, physics and self perception.
So, the fact that a GPT model can do all those things while being just a language model is extraordinary. It is an idea serializer. Written language is not the idea, is just the serialization of the idea, the human thought.
We serialize our ideas, and with a GPT we use the serialization to leverage some of the connections and behaviours of the idea, without actually understanding the meaning.
This makes both tools, language for humans and GPT for computers, a powerful abstraction that allows us to not need to think about every detail because some of the logic in the ideas is covered by the serialization that language performs over the idea.
I find all of this fascinating.
Add the other parts of the human brain (a GPT is only the Broca's area analogue) to have real logic and physics and mathematics understanding, and the predictions about AI will be true.