The Language of Control and other Semantic Problems

The English language is the de facto second language of the world.  And with respect to the many great contributions of science pioneers in all the many cultures of the world, English ends up as the language of engineering and control systems.  Not really surprising since English is also the primary language of commerce, even in China.

In most circumstances this would not be an issue.  However, as the world of automation progresses there are many issues hidden in our use of language used to describe technology.  The terminology we use to describe things in the world of technology has a great deal to do with our understanding, and misunderstanding of things around us.

A very controversial phrase these days is Artificial Intelligence.  It is not entirely artificial, and whether it is intelligent remains to be seen.   The most advanced software in the world on IBM’s Deep Blue, is in the end, written by humans.  As Watson’s inventor says, it is merely a “text search algorithm connected to a database”.  Watson’s virtue is not in it’s ‘intelligence’, it is the speed with which it can explore millions of options and come up with the right answer.

The same is true for the phrase “deep learning”.  There are many statistically significant relationships like correlation.  Deep Learning is merely a set of algorithms that test for significant relationships across sets of variables.  What makes this “deep learning” look like intelligence is really the brute force capability of the computer to make millions of calculation and determine which results are significant according to a set of rules that were written by human programmers.

So we use terms like learning and intelligence metaphorically, not literally.  These terms are NOT used literally and should never be interpreted as such.  But that’s what we do.  Which leads to the completely false notion that computers are capable of exhibiting human intelligence.  The only resemblance between computers and humans is the guy that programmed it.

Will computers ever “learn” in the way that humans do and acquire skill?  So far, this doesn’t appear very promising.  DARPA has been running the Grand Challenge of autonomous vehicle navigation for over a decade.  Thousands of programmers have spent thousands of man-years developing this capability.  Google, Tesla, Apple and other technology giants are spending billions of dollars and millions of man hours refining the work started by Darpa.

Is this a case where computers exhibit intelligence?  Not really.



Mechatronic Tips