Programs that can solve well-defined problems, such as telling a cat apart from a dog or win a game of chess, are called weak or narrow AI. Recently, computer programs have become rather good at this. Partially autonomous cars, recommendation systems, and image recognition are just a few examples that already exist in everyday use.
A program that could deal with a more complex environment and show human-like intelligence in multiple tasks – like riding a bike by the lake, looking at a view it considers beautiful, and deciding to paint it onto a canvas – would be called strong AI. This only currently exists in science fiction!
So, which one is more intelligent: a spam filter or a speech synthesizer? It’s a silly question, isn’t it? The filter and synthesizer represent narrow AI: they are only “intelligent” in one, very limited, field. Even though a speech synthesizer can produce speech comprehensible to humans, it will not understand what it has produced.