Intel's patented a variation on the Karatsuba algorithm. It hasn't simply patented 'multiplying two numbers' as stated!
Even though I have a couple of patents to my name I am generally against software patents, sometimes I think that for genuinely novel algorithms there should be an exception (e.g. RSA encryption). However, as there is clearly no way of objectively defining what a "novel" algorithm is (i.e. there is no algorithm for it) we can't really have this exception.
So yes, I would agree that algorithms should not be patentable.
Can't actually see the patent since the patent site's down.