Re: Can mutation and selection increase information?
Not even then. Remember, that was just an experiment comparing two different methods of selection. Any arbitrary sequence of any set of symbols would have served just as well in that experiment. A sequence of English words using Roman letters was just convenient for him and poetic as well. My MONKEY version of the experiement doesn't even use English words, just the letters in an arbitrary sequence (by default the alphabet in alphabetical order), though you could choose to have it use any string you want, even a sequence of words in any language you want (just don't expect support for special characters or diacritics).
Though there's something in Vlad's messages that looks suspicious. He keeps going on and on about "spontaneous evolution". I have absolutely no idea what he means by that, except that it appears to not use selection. Maybe it's meant to be creationism's bogus probability model that uses single-step selection (as discussed on my MONKEY pages) to make entire complex structures just fall all together in a single event by pure chance. Again, the probabilities of the two selection methods are discussed and analyzed mathematically on my MONKEY Probabilities page.
Of course, if Vlad means something completely different, then he needs to explain what "spontaneous evolution" is supposed to be.
ABE: The primary problem of trying to evolve actual language text or functional machine code (or even source code) is that those systems are not conducive to evolving because they are too brittle, they break too easily. Amino acid sequences for proteins do not suffer from that brittleness problem.
Thomas Ray discussed this brittleness problem in his documentation for his TIERRA simulation -- TIERRA Home Page.
Gotta leave right now. Running late for Carolina Shag class, followed by manning the door at a dance.