TL;DR: Creating information is almost always part of an act of communication, and a piece of communication has a context, purposes, a source and recipients, a channel, production/consumption/storage/delivery costs, plus other constraints.
All of this will condition the form of the information, and it's usually wise to consider this when creating information you want 'optimized'.
I usually wake up reading my e-mails and going through slashdot, and today's events catalyzed this post.
First, while reading a discussion thread on « Why Are We Still Writing Text-Based Code? » there was a funny quote (http://ask.slashdot.org/comments.pl?sid=4766093&cid=46192345
A picture is worth a thousand words.
A thousand pictures flipping past at 24 frames per second is worth ten words.
The article is about usage of graphical languages (eg. Simulink, LabVIEW, ...), which are not completely foreign to me, but I won't even be talking about that (while they are cool, and can be adapted to some applications, my main grief is that the most widely used are nightmares in terms of SCM, scaling and vendor lock-in).
The second event is that I got a work e-mail from a QA colleague that was describing a problem by sending 2 iPhone camera shots of a phenomenon. I can't go in the details but these images were mostly useless: the phenomenon was not clear, the quality was low (uncropped, bad white balance, bad focus) and they had been hand-edited to include a text pointer (so, I estimate that it took at least one minute to create, transfer, edit and include these pictures in the e-mail). The e-mail was totalling 7.63 MB, so it took me a while to download it from home. A simple sentence would have been sufficient and clearer to describe this problem. Oh, and of course, the company has an IT system called an issue tracker to deal with this...
I have a serious problem not only with wasted bytes, but even more regarding valuable information that is not 'optimized': not concise, or using a lesser appropriate format.
Text has the nice habit of being very easily searchable, editable, it can be consolidated easily. Information in textual form is very convenient and forgiving. It's really hard to write a piece of text that is completely useless to the recipients (but it definitely does happen).
Pictures are helpful for an instant comprehension of a phenomenon, or can convey feelings, but unless they are art, they are mostly a liability unless temporary.
Generally pictures are quite costly because the amount of information per byte is quite low, but there are cases where it's hard to avoid them (I'm currently dealing with terabytes of raw photographic data in my job), and there are also exceptions.
Schemas (graphs, flowcharts, visualizations) are a specialized form of pictures which is usually quite useful, and compact.
Doing serious schema works will usually lead you to using programming languages to write the schema.
Looking back for information in a picture is costly.
Motion pictures, are in my opinion falling in the categories of either works of art, memories, or "temporary" (waiting to be consolidated in another format, or deleted).
Parenthesis: It's plaguing the internet and I really hate it when some important piece of technical information is only available in the middle of a video-recorded presentation, an ungrepable waste of both bytes and time. I find that borderline disrespectful for the audience. Of course, at some point we're all presentation newbies.
Whether you want to share information, make sense of data, do indexing or data archeology, you will find that introducing non-text data with text meta-data is useful, be it a pointer, an outline, a summary. Do that and it will pay off sooner than later.