KNOWLEDGE AND DATA

Tom Gruber is head of advanced development for Siri at Apple. He has been at the vanguard of AI (Artificial Intelligence) from the very first. And he has specifically been interested in speech recognition by computers. His work in the early 1970s established what is called the Hearsay project, which essentially is Siri in embryonic form.

Early on, there were two schools of thought regarding how to get machines to think.  One was logic driven, and the other was data driven.  The logic approach tries to get machines to think using the basic building blocks of thought, similar to the components of a sentence.

The data driven approach postulates that knowledge is really only a matter of making connections.  So if you have enough data, it will only be a matter of time before correct matches produce the right answer.

“Gruber was in the logic camp, and the approach is ‘no longer fashionable.  Today, people want no knowledge but lots of data and machine learning.'” p 233

And of course there is a deficiency in the data driven approach.  “No one really has any idea of what the models know or what they mean.”  Immense databases end up mimicking “how we hear and see, but not how we think.”  So AI, in it’s current form, is simply a matter of disconnected data.  It is not thinking.

When we draw representationally, we mimick how we see, but not how we think.  When we draw, we want knowledge.  We don’t want data for data’s sake.  This AI example illustrates that the distinction between perception (data)  and knowledge (logic)  is crucial.  The logic driven approach is holistic.  This gets to the heart of why drawing is so important.  Transparent Drawing requires knowledge.  It requires logic and thinking to understand how something works holistically.

1.  Merchant, Brian. The One Device. Little Brown and Company. New York. 2017.