The death of writing, part 1

Preface: This is NOT a boo-hoo post about how people watch too much television and can’t spell.

I think of myself fundamentally as a writer. I love to read, I like books, I like magazines. On, I ignore the video clips and read the articles. The written word is for me the fundamental and primary way to exchange information.

Now it’s not that difficult to imagine a future in which writing is used mostly for labeling. Video and audio will be fundamental and primary, and writing will be used for short-form annotation of the video and audio. Writing will be metadata. Labeling.

In the past video and audio weren’t used in this way so much because of limitations of (obviously) transmission and storage, but also less obviously of editability and searchability.

The first two issues have largely been solved: broadband, mobile, gigabyte hard drives and flash memory. Internet, cable modems, satellite, Youtube, iPads, Android.

The third issue, editing, is mostly solved when you think about simple tasks; video and audio editing software comes standard on lots of computers now. Not everybody wants to develop Final Cut Pro skills, but it’s available. “Editing” encompasses more than that, though. If you are trying to organize and convey a large amount of information, or update existing information without reshooting the whole video, writing still offers advantages in speed and flexibility. In fact, writing in the computer age is so flexible that I often write before I organize. I start putting down various thoughts and then gradually re-order them as I go. Sometimes I stop and do an outline AFTER I’m halfway done writing. I don’t think this was an efficient approach in the typewriter era.

The fourth issue, search, is where it gets really interesting. Right now, automated information retrieval is text-based– think Google. But it doesn’t have to stay that way.

Dragon Naturally Speaking is voice-recognition software. It converts audio into text. It’s very limited even after decades as a commercial product. We have tried it out for interviews – can it transcribe automatically while the interview takes place? It can’t. It has to be “trained” to recognize a speaker’s words correctly. You have to work with it to get accurate output of your own voice, let alone some random soprano interviewee with a Scottish accent and a mild stutter.

BUT the people who make the software say that’s a limitation imposed by computing power. They say as computers get faster and faster, this barrier will inevitably fall. And that means that audio-based search is also inevitable. Imagine: You have 200 voice mails saved on your iPhone, as nobody uses email anymore. You need to find the ones about the next budget meeting. You say to the phone: “Search mail – keyword budget.” Up pop the six messages that you audiotagged “budget”, and two others in which the word budget was used during a conversation.

Here’s where the written word appears in the scenario: the list. It’s metadata. Labeling. Just like I said. Because you the human can scan the list faster than it can be read aloud. (Note here a distinction between scanning – visually processing a list of metadata, which you do very quickly – versus searching through a large corpus of data to find all objects matching a specified pattern, which the computer does very quickly.)

Anyway, you look at the list and say “play first David” and hear the top audio message from David. You say “skip – keyword supplies” and the audio track skips ahead to the part of about purchasing supplies. And so on.

This is all possible once the processing power is in place to convert your voice commands and to search the audio files at a high speed to find the right matches.

Same with video. Surveillance companies now embed “video analytics” in their products. You can, for example, have a camera in a parking garage that doesn’t stream video to the guard office until it 1) detects motion and 2) recognizes that the motion is cause by a person (as automatically distinguished from a mouse, or a gust of wind). With enough horsepower, you’ll be able to tell your phone, “find Fred tae kwon do” and it will bring up all the videos of your son’s TKD practices, or the clips on the instructor’s website showing how to correctly do the Tae Guck Il Jang form. Or search the entire web for semantic constructions like “find footage of crocodile attacks”.

I wrote about some advanced application of video analytics almost four years ago for my website/magazine in The age of analytics , noting how massive computing power changes the game – there’s a chess connection, for my chess junkie friends.

Now, ironically, I see the broader implications. Maybe my next “age of analytics” and “death of writing” pieces will be on YouTube.


12 thoughts on “The death of writing, part 1

  1. This is what I want to know: Has anyone thought about what would happen if, for some reason — oh, I don’t know, maybe a nuclear bomb goes off in the middle of Manhattan — the Earth/U.S./Europe/Asia/ lost its electrical power for a sustained period (let’s say three months)? What the F would happen? What would people do? All of their books/email/voice-mail/work/GPS/music/video/entertainment/ every damn thing people rely on would be gone, lost. We rely too much on computers. Take my word: Some day 300 years (or sooner) from now we’ll be screwed, because everything has gone digital and they’ll be no power to charge the world.

    Don’t throw out your old books. You’ll need something to read when the power dies.

  2. I knew I could count on you, Howard :)

    The thing is all electrical devices will be individually wind/sun/human-powered by then. The grid will only exist as a backup/exchange mechanism.

  3. Books are like walking. We have new ways of transportation that are faster, but our legs will always have a purpose.

    On, I ignore the video clips and read the articles.

    Hear, hear. I hate it when they play the videos automatically, and the talking heads overpower music.

    I’ll actually read the whole article when there’s more time. Probably could have done that in the time it took to write this comment.

  4. Wrong crowd, wrong crowd….

    The book is really going to die the day Google retails its super-fast book scanner. When that happens, people will swap books like MP3s. …and that’s the day my entire chess library becomes worthless. (I knew I shouldn’t have bought so many….fool was I to think I could resell them.)

  5. On voice recognition software: This is actually one area I’m not sure we’ll see progress soon. This opinion is based, of course, on anecdotal evidence from a small set of articles.

    The gist is that despite significant increases in processing power, voice recognition software hasn’t improved much in the last 10-15 or so years.

    Maybe there are some high-level programs that have shown more progress.

  6. I wonder if they could pre-build a wide array of “voice profiles” – maybe they already do this – in an effort to shortcut some of the training time. I’m calling someone in India, someone in Scotland, someone in Atlanta, I click on the right box and get closer to reasonable recognition right off the bat.

    Just a layman throwing storage/bandwidth/processing power at the problem here.

  7. That is a possibility.

    It seems that voice recognition does well when the set of words is limited. For example, phone menus, where the system might have to recognize “yes”, “no”, and numbers.

    Detecting “one” is easy under these conditions. The general case has to distinguish between “one”, “wan”, “on”, “wane”, or any such set of similar words.

    I also noticed this month’s Chess Life mentioned “Masters of Technique”, a book edited by one Howard Goldowsky. Could this possibly be the same Howard that frequents this site?

    1. “Our results suggest that male chess players choose significantly riskier strategies when playing against an attractive female opponent, even though this does not
      improve their performance.

      Women’s strategies are not affected by the attractiveness of the opponent.”

      Good find, Howard.

Comments are closed.