A question worth answering properly.
THE SHORT ANSWER
Not in the way the word usually means.
These systems perform tasks associated with intelligence without anything resembling understanding.
WHAT THEY DEMONSTRABLY DO
Produce text statistically consistent with their training and your input.
That is sufficient to be useful. It is not intelligence in any ordinary sense.
HOW THEY FAIL
In ways that understanding would prevent. Confidently asserting contradictions, inventing citations, missing obvious implications.
A system that understood would not make those errors.
WHY THE QUESTION MATTERS PRACTICALLY
If you believe it is intelligent, you trust it inappropriately.
If you treat it as a very capable pattern-matcher, you use it well.
WHY IT SEEMS INTELLIGENT
It was trained on text written by people who were. The resemblance is a property of the output, not evidence about the system.
THE USEFUL POSITION
Judge it by what it reliably does, not by what it seems to be.
