Episode 13 Pattern Matching

  • Posted on: 19 September 2017
  • By: C.J.

A recent claim of deciphering the Voynich Manuscript and a heavy dose of superstition powers this podcast!

iTunes Icon Image: 
The logo for the podcast Self Help for Robots
iTunes Episode Subtitle: 
Welcome to the lucky thirteenth podcast: Pattern Matching!
iTunes Episode Summary: 
Welcome to the lucky thirteenth podcast: Pattern Matching!


C.J.'s picture

Welcome to Self Help for Robots. I'm your host, C.J. Pitchford, and this is the thirteenth episode I’d like to call “Pattern Matching” but first, what’s in a name? A name like “Voynich?” Wait, who?

The Voynich Manuscript has been called the most cryptic example of its kind—because of its handwritten text that defied translation—but it may have just been deciphered by Nicholas Gibbs for the Times Literary Supplement. Mr. Gibbs detailed similarities among popular medieval medical manuscripts and the contents of the Voynich Manuscript. You can follow the link for more from the selfhelp4robots.com website!

Apparently, once Mr. Gibbs detailed the use of Latin contractions in ligatures—the popular ampersand is a ligature of the letters of the latin word “et” that translates to “and”—while also maintaining that every letter is an abbreviation of a word, the text apparently becomes recognizable as common advice by Galen, Pliny, and Hippocrates.

At least, in the short example provided at the Times.

But that’s part of the problem. Us non-robots are so good at pattern matching, we don’t even consider it special. People really don’t need much to go on before they make deductions, inferences and leaps of logic based on a supposed pattern.

Until, that is, they try to teach a computer how to find similarities. That’s when the assumptions drop out from under your feet, or start to fly out the window.

You know, the first pattern matching concept I can remember is the song “One of these things doesn’t belong…” But first a robot needs to be taught what a pattern is, so it can’t even do that until it knows how to look for.

So, to try that with artificial intelligence, you first have to give a robot a way to adapt and augment it’s own data and models. A face doesn’t always have two eyes and a mouth visible, but once learned, the pattern persists even when truncated or hidden. Facial recognition systems are getting better, but they are still fairly easy to trick.

Remember the manuscript I mentioned? The reaction to Mr. Gibbs’ proposal from the research community of professionals and amateurs alike is that his facile explanation is insufficient to explain the encrypted text, as the decrypted versions aren’t grammatical and in some cases, appear nonsensical.

It’s possible that Mr. Gibbs is seeing patterns where they don’t truly exist. Machine Learning as a discipline tries to ameliorate the process by using sheer numbers. The more you look at the patterns—the thinking goes—the better you become at avoiding false positive matches.

But just because machine learning can predict when I will get cancer doesn’t mean that there are other ways to learn that from the same information. The data—my cells and their functions—are already there, and there are other methods to detect when I will contract a disease. Machine Learning can only tell us what we already know—or, ought to already know if we looked.

Yet, in order to learn something completely new, we first have to program the basis for knowledge. Instead of linear processing, now being pushed to its limits in machine learning, we can expand the scope of processing by examining non-linear, non-binary, even irrational approaches.

Once a non-robot like me is exposed to a pattern, other concepts are routinely evaluated in comparison to that pattern. Like the song and game I mentioned above, the ability to see patterns is assumed. The struggle for Machine Learning isn’t in finding something that doesn’t fit the pattern, but it’s how to create new patterns.

If non-robots are seeing patterns even where they don’t exist, while robots struggle to even develop the ability to match patterns as effectively, we need to re-think our approaches to the entire problem!

Remember, if you’re struggling with depression or anxiety or any irrational feelings that are bothering you, please seek out a mental health professional that you trust!

As always, robots and non-robots, keep helping yourself.

C.J. Pitchford, Paracounselor