To me, it is interesting to note that "intelligent" behavior (I suppose defined as the intentional manipulation of the environment for a specific end result, etc.) does not arise in a totally linear fashion down the evolutionary tree leading to us. Instead, it appears in creatures such as octopi, birds and other species separated by millions of years of evolution.

So, my question is, are there common selective pressures occuring in nature that push these unrelated species toward intelligence? If so, what is it?

On the other hand, I don't think it takes a rocket scientist -- or a biological scientist (har har) -- to see that there is a somewhat linear facet to the evolution of intelligence -- am I right? Humans are smarter than apes*, most apes smarter than most mammals, mammals generally are more intelligent than reptiles, reptiles more so than fish, and you see what I'm getting at.

I suppose this is the sort of thinking that leads to the myth that evolution is "progressive." But the point is, that sort of pattern needs an explanation. Why would the evolution of intelligence not be more widespread and diverse?

*Politicians aside.

There are actually several great questions in there! I will try to tackle the last one first...
Intelligence (or at least what we think of as intelligence) is not more widespread than it is probably because it is not always adaptive. A big brain takes up a lot of space, uses up a lot of energy (more than any other organ, pound for pound) and takes a long time to grow. The human brain takes at least twenty years to fully mature. Also, whereas a big brain allows for complex behavior, you can have complex behavior without a big brain. Take dung beetles for example: the act of gathering together a ball of dung larger than oneself and rolling it over long distances to a burrow is a pretty complex behavior when you consider all of the steps involved. But this behavior is purely instinctive and, therefore not really intelligent.
As for the selective pressure in favor of intelligence; natural selection usually produces species that are adapted to their current environment. If the environment changes, a species can 'adapt' in the sense that inherited traits that are favorable in the new environment will gradually predominate. However the process takes many generations.
The main advantage of an intelligent brain, however, is the ability to adapt quickly to a changing environment. To go back to the dung beetles again, if there are no droppings around for them to collect, they are in big trouble because they don't have any other way of getting food. An intelligent species, however, would be able to learn new ways of finding food.
So, the assumption is that intelligence evolves in situations where food or some other resource is scarce or unreliable such that the advantages of being able to adapt quickly outweighs the energy cost and vulnerability associated with having a big brain.

One theory about the rise of intelligence in humans specifically is a sexual selection effect, where mates were preferred over others who were able to exhibit some as yet undetermined intelligent action. Art has been proposed to be a mechanism, or perhaps an bility to make better tools. Sexual selection is a funny thing and produces outlandish, counter-intuitive adaptations. It could be that humans represent a line of apes who were consistantly selected for intelligence by this method, though it's almost certainly not that simple.

But like Brent said, intelligence comes at a metabolic cost. In humans it's not just brain size and length of maturation that are problems, but earlier births and increased post natal parental investment also place significant limitations on breeding. The same can be said of any organism, especially ones where the young are essentially left to fend for themselves and who therefore have greater benefit from "hard wired" behaviours rather then intelligently discovered behaviours.

Just a quick note on the issue of linearity in the evolution of intelligence. What may seem to be linear relationship when looking from the present backwards in time, is of course a property of the frame of reference. Rather like solving a maze by starting in the middle and working outwards - the confusing branches that lead elsewhere are easily ignored (or not even seen).

Intelligence does require a fairly complex neural system to be established, which arises from the development of more simple neural systems. As a result, when looking from the present backwards we see that brains get progressively smaller and less complex as we move through time.

Of course, we then reverse our perspective and see that brains get bigger and more complex over time as we follow the same route back to the present. When we do this we miss all of the branches where brain size and complexity stabilises or becomes reduced. In addition, we may also miss instances where brain size reduces and then increases again in a lineage, simply because the fossil record and species sampling are both imperfect.

So although brain size looks like it increases linearly, that is partly an artefact of the requirement for progressive levels of complexity of the infrastructure that supports intelligence and it's partly our perspective which favours looking at intelligence as an optimal endpoint rather than just another adaptation that can vary depending on its selective value.