Expert systems are really a subset of AI as are learning algorithms. Learning algorithms may be "cooler" AI but I'd guess that people get more done with expert systems.

Speaking of learning systems, it would be quite easy to change this program to utilize a decision tree (see ISBN 0070428077, chap. 3 for a good reference).

To do this, one would need to create a set of examples of the form
Dolphin 1 0 1 0 0 0
although I'd add some more animals and attributes to make this interesting. You need a number of examples per animal. The dtree algorithm uses the examples to build a tree and decides which attributes are most important and moves those attributes towards the top of the tree. It might place the same question in more than one place and might ignore questions in certain subtrees. For example, knowing that an animal gets stuck in tuna nets narrows the field down quite a bit.

Of course, DTree's tend to be used for more broad classifications, like deciding the party of a senator based on her voting record.