First off, let me say that the current batch of Artificial Intelligence models are not that exactly.
They are a neural network with a large language model that is able to scan the internet at rapid speed and collate the whole of the human experience (on the internet, with all it's flaws).
So you're not getting true intelligence, just through training the model which of it's assumptions are correct you train the neural net to be more correct than not. Eventually.
But it's interesting that such a basic premise: sum up the whole human experience online and offer answers to questions posed based on that gives you such a powerful tool.
However, it's quite clear that where the online information is missing, the various AIs crash and burn.
I'm not telling which area I posed questions to AIs, but I know for a fact that there is scarce information online about it. Every AI failed miserably to respond with any sort of response.
The fact that people are considering such flawed engines should be involved in informing government how to perform it's task is to me hugely concerning.
It is in no way equipped to deal with the variations of life, especially the unknowns that require a very alternative thought process to resolve.
Currently what is known as A.I. relies on the reserve of human knowledge in order to inform any questions put to it. If you start asking about abstract ideas, it gives abstract or vague responses, or almost directly quotes philosophers. It does not decide on it's own what the future may be like, it parrots to majority viewpoint gleaned from the net.
And if this can outthink the thinkers, I hate to think what sort of thinkers we have currently. Pretty shallow ones, it seems.