I favor a model of intelligence where one measures problem-solving capability along multiple axes and defines an N-dimensional shape that describes a hypervolume of 'general' intelligence. One could certainly argue that because this program solves problems that can only meaningfully be measured along one axis that it has too few dimensions to be called intelligence. The trouble with that is software generally works best when specialized for problems with a single axis, so by that definition 'general intelligence' is either unlikely to happen in that space at all or won't be energy efficient if it does.
What would likely have a higher intelligence is a 'cyborg' - which could just be a human accessing this program via conventional means - who adds this program's problem solving capability along one axis to their existing hypervolume of intelligence. So maybe not artificial intelligence, but definitely an artificial intelligence expander.