Machine learning for the people

It began with a good intention. Someone, somewhere, wanted to create an autocorrect function that improved both spelling and grammar, above and beyond the makeshift implementations found traditional writing programs. After a number of false starts, it finally bore fruit, resulting in a startlingly effective piece of software that could make just about anyone appear a fully literate, articulate person in command of the written word.

At first, this increased clarity of communication across the board. Common misunderstandings decreased substantially, irritation over the intricacies of interjections and prepositions melted away, and overall both sides of written communication became that much smoother. Their, they’re and there – for a while, it was a solved problem.

But then

A few years later, a backlash emerged. Books, blog posts and other pieces of writing all started to blend together into a universal soup of common language. It was subtle at first, but when a prominent social media icon pointed out that a heartfelt recollection of young love lost read exactly the same as a summary of the fluctuations of the stock markets, it became clear as day. The algorithm did enhance legibility, but it was only ever an algorithm.

Needless to say, writers who wrote imperfect prose rose in popularity during this period. In response to this, companies providing proofreading software began to introduce small flaws in their programming, subtle enough not to be noticeable in short pieces of text, but sufficient to differentiate longer works from each other. This dynamic kept up for quite some time, with readers becoming ever more adept at pattern recognition and the programs becoming ever more subtle in their introduction of errors. Readers and writing tools, locked in an ever escalating arms race. In the end, it turned out that the quality of writing generally improved from not using any such proofreading software whatsoever.

Out of the algorithmically imposed lingua franca came a resurgence of antediluvian atavisms, such as unrelenting human editors who simply would not accept you doing less than you were capable of. Humanity, battle-scarred and not quite certain about what language even is any more, resignedly accepted this return of the once dreaded red pens.

%d bloggers like this: