[wordup] Mind the Scaffolding

[Adam Shand]

Neither the article in the LA Times or the Google cache of it appears to
exist anymore.  Regardless it was mostly this analysis that is
interesting.  You have to get right down to the bottom for the good stuff.

From: http://www.hyperorg.com/backissues/joho-aug15-01.html#mind

Mind the Scaffolding

Frank Schmidt calls to our attention an article by Michael Hiltzik in the
Los Angeles Times about Cyc (pronounced "sike," as in "ensikelopedia"),
Douglas Lenat's gargantuan attempt to misunderstand human thought.
Seventeen years in the making, Cyc thinks that we're intelligent because
we have an internal database of commonsense facts. According to the
article:

  The project already has consumed an estimated 500 person-years and $50
  million in investments from, among others, the Defense Department, the
  pharmaceuticals company GlaxoSmithKline, and Microsoft co-founder Paul
  Allen.

  ...

  The system today encompasses more than 1.4 million assertions--hundreds
  of thousands of root words, names, descriptions, abstract concepts, and
  a method of making inferences that allows the system to understand that,
  for example, a piece of wood can be smashed into smaller pieces of wood,
  but a table can't be smashed into a pile of smaller tables...

And what do we get from this massive investment:

  In one recent demonstration for a Defense Department project, a Cycorp
  engineer informed the system they would be discussing anthrax. Cyc
  responded: "Do you mean Anthrax (the heavy metal band), anthrax (the
  bacterium), or anthrax (the disease)?" Asked to comment on the
  bacterium's toxicity to people, it replied: "I assume you mean people
  (homo sapiens).  The following would not make sense: People Magazine."

Talk about your artificial stupidity! 500 person-years to get the thing to
not know the difference between a bacterium and a heavy metal band without
asking!

By coincidence, Cyc is the opening example in the book Being There by Andy
Clark which I've been reading and was going to recommend even before Cyc's
PR machine decided it was time to start pumping. Being There is brilliant.
In fact, the second half is too brilliant for me: it goes down some
capillaries of thought about the brain that lose me. But that's ok. The
book has one central thought that is so obvious that it seems like we must
have known it all along. Clark doesn't claim credit for the idea, but he's
pulled together several lines of thought, extended them and expressed it
in a way all his own.

Attempts at AI such as Cyc play right into what I've started calling our
"default philosophy" in my book-in-progress. Thanks to a couple of
thousands years of thought, we are now perfectly at home with the idea
that our mind creates internal representations of the external world. We
manipulate those representations and decide on actions that then affect
the external world. Cyc attempts to make explicit our internal
representation of the world. But, that isn't in fact how we work. I've
argued for years, fruitlessly, against this representational model by
pointing out that it fails to describe how we actually experience the
world, but this "phenomenological" approach inevitably draws the response
that the causes of our experience may not be like our experience -- a
position that I find maddeningly obtuse, but one I have never managed to
subvert. Clark takes a much more effective, non-phenomenological path. He
says that the brain is a literal neural net. Neural nets enable us to
explain how we do things like catch a Frisbee without having to have an
inner map of the space the Frisbee is traversing and without having to
calculate complex arcs of motion: we learn to associate the position of
the Frisbee in our visual field with the movement of our hand. Fine and
dandy, but, Clark asks, how can a neural net explain our abstract
thinking?

He starts with the example of doing complex multiplication problems. Our
associative neural net masters the times tables (or "math facts" as the
schools now say). We then learn a technique that lets us conquer a tough
multiplication problem by repeatedly applying the times table. The
technique involves writing the numbers on a piece of paper, drawing a line
underneath them, etc. Clark's point is that our mind's ability to do the
requisite abstract thinking depends on the paper and pencil. Abstract
thought requires "external scaffolding" -- Clark's term for stuff in the
world that we can alter in order to help our neural net do tough, complex,
abstruse reasoning. This external scaffolding includes the simple ways we
organize the world (e.g., alphabetizing our CDs), the chalkboard the
physics professor uses, libraries, the Web, language itself and social
institutions. Without these, we are naught but damn dirty apes. The
characteristics of our mind that we identify as most peculiarly human
depend on our ability to alter our world to help us think. Our human mind
is inextricably bound up with the world and its manipulation.

I find this thrilling. It redefines mind in a way that busts us out of our
self-imposed mental prison so we no longer have to think that our minds
are primarily individual and isolated from the world. Quite the contrary.
Our minds are not only formed by our culture, they would be impossible
without the things of our culture. And as we rapidly increase our
"external scaffolding" -- for example, with the growth of the Web -- we
are expanding not only our "information resources" but our minds
themselves.