The Deletist

I’ve noticed that as a species, we seem to give a lot of credit to computers and technology for all of their marvelous capabilities.  Often we do this without realizing that humans are responsible for all the design, development, and programming.

In reality, computers really only do what we tell them to do.  And even more importantly, computers don’t do nuance.  Basically, a computers can do anything, as long as it’s able to be translated into the computer’s language.  Essentially, everything must be distilled down to a binary decision: Yes/No, True/False, 0/1, A/B, black/white, etc.

I consider this factor often when doing backend system design for clients.  In order for technology to take over for a human (e.g. by automating a process), all of the decision points must be simplified to computer language.  This can be tricky when a particular process contains too many factors, or “gray areas”, to result in a binary decision point.  For example, consider a college application.  The computer would be able to make decisions based on GPA and test scores, but evaluating the candidate based on other, more nuanced factors, such as extracurricular activities or the essay portion, must be handled by a human.

Another area where computers lag behind humans is in facial recognition.  Overall humans are able to identify the same person in photos or video more reliably and accurately than computers can.  Although computers have improved in this area and can accomplish some pretty amazing things in terms of auto-classifying photos, they still have a long way to go.  The other day a friend of mine was showing me some of his childhood pictures.  I’ve only known him for less than a year, but I was immediately able to pick him out of his primary school class photos.  I doubt a computer would be able to do that yet.

A lot of advances have been made in AI (Artificial Intelligence) lately that may one day allow computers to think outside of their programming and to better anticipate our needs.  However, based on how poorly the prediction and auto correct works with my text messages, I’d say they still have a long way to go.  So while we’re eager to give computers (and technology) a lot of credit for all the amazing things they can, and will, do, there’s still a place for human brains.

I typically use swype writing to create messages on my smarphone.  Swype writing [link here] allows me to enter words by moving my finger continuously across the keyboard over the appropriate letters.  For example, to type Monday, I start with my finger on the “M”, slide up to the “o”, down to the “n”, over to the “d” and “a” before finishing on “y”.  The word is created without lifting up my finger to input each letter individually.

Often the smartphone figures out what the word is and suggests a term before I’ve finished “typing” it based on predictive text.  After completing each word, the predictive text feature also suggests words to follow based on my usage patterns.  For example, if I type “What”, I am automatically offered options for the next word such as “you, time, is, are, about, time are”.  Sometimes I can complete the message by selecting the offered options.  Most of the time this works well.

Sometimes, however, the smartphone tries to be too helpful.  I turned off the auto-replace feature on my smartphone because it kept inserting email addresses into my messages based on a few letters, among some other bizarre things happening.  It was so annoying!  At times, the prediction and auto-replace are so wrong, and not based on my language patterns, I can’t help but think some developer had a good laugh while doing the programming.  Here are some examples.

“And”, a common and frequently used word, was almost always auto-replaced as “abs”.  After turning off auto-replaced, I’m now offered “abdominals” as an option when I insert “and”.  I’m not a fitness trainer.  It’s rare for me to use “abs” or “abdominals” in a sentence.  When would these words ever be more likely, or common, than “and”?

About 85% of the time “also” was automatically replaced with “Akzo”.  Huh?  What is Akzo?  Also puzzling is why “also”, another common term, would be replaced with an obscure proper noun.  It just doesn’t make sense.

Another strange correction is “with” frequently turning into “wuthering”.  I was a literature major in university and I’m 100% sure that I’ve never used this word in a sentence, unless I was referring to the book, Wuthering Heights.

I understand some training and customization is necessary, but some options are so off that I wonder how they even got programmed in.  It’s made proofreading, even for trivial messages, essential.