My opinion? I think the term “artificial intelligence” is misleading as it is still a mathematical model. In a simple test of a game of tic-tac-toe (a game that usually ends in a draw), I beat all of the best-performing AI models by systematically playing the same sequence of moves. And they were completely unable to adjust their strategy even when given the strategy I was using.
In a second test, based on words intended to represent their strength, they were unable to check the application of two simple rules of a word order game: the first letter of the given word must be identical to the last letter of the previous word and the given word may still be never been stated before.
So if you don't anticipate all possible cases and train them on them as soon as they leave their data set, it's ruined. And since technological, energy, infrastructural and climatic limitations make it impossible to specify all possible real-world cases, this “*AI*” will remain an approximation and simulation of the intelligence of even a mouse. Unless they specialize in a specific and limited area.
Maybe a new technology seems to be able to really learn and really understand the context and reasoning. But it is certainly not a technology based on binary and silicon, since the world is analog and not digital.