I didn’t write this blog post. It’s me. It’s my words. But I didn’t write it.
I guess it all starts back with SpellCheck. Back in the dark ages of the early computer days, you wrote your text and then asked the program to compare it to a dictionary file. It was better with typos than with homonyms, but you learned to read for their/they’re/there. Meanwhile, while the program caught your misteaks.
Then the programs were written to point out your errors as you went along.
Then your words got
fxied fixed for you as you typed. First it was an option. Then it was the default that you had to actively turn off.
The next step was for The Powers That Be to add auto-complete for words. I see that you have typed xy, do you want xyz? At first you had to say, ‘Why, yes, that is what I meant, thank you.’ Over time, this became became harder and harder to avoid. You had to go farther and farther down the settings menu to say, ‘No. Please stop. Do not finish my words for me.’
And then there was grammar. ‘The quick brown fox
were was …’ Who doesn’t need the occasional help with verb tenses?
With auto-grammar came auto-complete for phrases. ‘Have a nice … day’ was harmless enough.
Then it got personal. The program began sticking in ‘But I digress’ at the end of paragraphs where I had wandered from the topic. I mean, yes, I do say that. Fairly often. But most people don’t.
The algorithm was clearly tuning itself to me.
Then one day, the suggested completion ran to several sentences. The rest of a paragraph, in fact. Yeah, that sounds like me.
Where did that come from?
Am I that predictable? Well, I’ve written over 3,000 posts on a small segment of a niche topic. A certain amount of repetition could be expected. It only seemed to happen with blog posts, not work files. But then, blog post files had titles along the lines of ‘draft year month day title’. Not hard to collate those from my directory.
Are we that predictable? Humans are capable of epic leaps of the imagination. Occasionally. More often, we are pondering how best to wade through the next 24 hours. Have you ever asked someone a question knowing what the answer will be? It’s like that. For most of us. Most of the time. Given enough data.
Then auto-suggest turn into auto-complete.
It tried to turn it off. The best I could get was having the program highlight the suggested sections so that I could backspace over them. I started to let them stand more often than not.
The program remembered my subtitles when I didn’t. It linked to other parts of the blog, as is my wont. It found interesting content for my introductory section on the outside world.
My blog numbers began to creep up. Nothing dramatic. Just a nice, pleasant upward trend on who came to visit and how long they stayed. The posts sounded like me. Or, to be honest, they sounded like me on a good day. My words. My phrases. But mixed. Re-ordered. Shown to their best advantage. What I would have produced if the writing muse had showered me with fairy dust.
The program was even willing to make jokes about our robot overloads. After all. that’s exactly the sort of thing I would have said. Although, maybe it’s not a program. Maybe it’s a person on the other end compiling my back posts and predicting what I will say.
Which would you rather have? The AI from Cat Pictures Please (Kritzler Clarkesworld 2015) or a human who gathers data as an experiment to mess with your head?
Stay safe. Stay sane.