The turing test isn't very useful. At its core, it merely tests whether a computational agent can imitate a humnan agent well enough to fool an typical individual.
Turing did not mention a "typical individual". The question is whether it's possible to make AI which is indistinguishable from a human. Obviously, it makes sense if an interrogator comes prepared if we want to test this.
We had AI indistinguishable from a bored teen who isn't really into the conversation decades ago. It's also really easy to pass a turing test if your model is a pissed off friend who won't respond to your texts.
Well, Turing explicitly formulated it as a _game_, and for a game to be meaningful, players have to understand rules and have a desire to win. And given that the question was "Can machines think?", a human playing the game should have a good thinking ability.
Game with bored, disinterested players would be entirely meaningless.
It's looking more like "how can I tell if I'm talking to a machine unless the potential person on the other end is strapped into a torture chair and forced to respond rationally to my inane philosophical ramblings."
And I'd like it more as a Gedankenexperiment if people weren't talking about it as a tool or metric. That kind of thinking gains momentum.
There's nothing in the turing test or systems that solve it that are really amazing progress. To me they are just an expected and obvious outcome of the general improvement of scientific modelling of reality.
Here's something that I'm not blase about: AlphaFold. That is one of the crowning achievments of humanity. It solved a problem that people have been working hard on for 40 years using an algorithm that is less than 5 years old on a computational framework that's a couple years old, on hardware with ML training powers orders of magnitude higher than anything that existed 5 years ago, and conclusively demonstrated that evolutionary sequence information, rather than physical modelling, is sufficient to predict nearly all protein structures. And, once the best competitor had a hint how it was done, they were able to reproduce (much of) the work in less than a year.
Now that's amazing. World-class. Nobel-prize worthy. Totally unexpected for at least another 10 years, if ever. Completely resets the landscape of expectations for anybody doing biological modelling. However, it also won't transform drug discovery any time soon.
Which tells us nothing at all.