Dealers of Lightning (22 page)

Nothing of the kind existed in the computer world in the 1960s.
Machines differed in shape, size, and architecture down to the cir­cuitry inside their cabinets and the sequences of digital ones and zeros
delivering instructions to the central processing unit. The same eight-
bit sequence, say "11110000," might tell a Burroughs computer to add
two numbers together and a Control Data 6600 to divide one by the
other. Each machine had its unique method for everything from stor­ing files on disk or drum to performing basic mathematical functions.
The differences were entirely arbitrary, no more consistent than if the
pedal by the right foot operated the accelerator on a Ford but the
headlights on a Chevy.

Nor did the manufacturers see any advantage to marketing machines
even remotely like their competitors'. Once IBM sold a system to
United Airlines it could rest assured that the frightful effort of rewrit­ing software, retraining staff, and moving tons of iron and steel cabi­nets around would make United think very long and hard before
replacing its IBM system by one made by, say, Honeywell.

Therefore Kay, who had programmed everything from a Burroughs
5000 at the Air Force Air Training Command to a Control Data 6600 at
NCAR, the National Center for Atmospheric Research, was compelled
to become a student of computer architectures. Subconsciously his mind
was absorbing the principles of programming that would grow a few
years hence at PARC into an extraordinary advance in software design.
As he recalled later, however, at the moment "I barely saw it."

So too did he assimilate only subconsciously an article in a technical
magazine he came upon while debugging NCAR's giant CDC 6600 in
Chippewa Falls, Wisconsin, in 1965. The magazine was
Electronics.
For
its thirty-fifth anniversary issue it had invited a few industry leaders to
plot a technology curve for the next ten years. The research director at
Fairchild Semiconductor Co., a brilliant engineer named Gordon
Moore, contributed a four-page piece insouciantly entitled "Cramming
More Components onto Integrated Circuits." The essay forecast that as
circuits became more densely packed with microscopic transistors, com­puting power would exponentially increase in performance and diminish
in cost over the years. Moore contended that this trend could be pre­dicted mathematically, so that memory costing $500,000 in 1965 would
come all the way down to $3,000 by 1985—an insight so basic to the sub­sequent growth and expansion of the computer industry that ever since
then it has been known as "Moore's Law."

That day in 1965, however, Alan Kay skimmed Moore's article and laid
it aside, unmoved. The dream of a computer scaled down to serve a sin­gle human being would not come to him for another couple of years. As
he toiled in Chippewa Falls on a room-sized, freon-cooled CDC 6600,

Gordon Moore's astonishing prediction that electronics had embarked
on a journey of unceasing miniaturization seemed to have no relevance to
his life at all.

"I was in an embryonic state. I didn't want to work and get a real job,
but go to graduate school. The only criterion was that it had to be
above four thousand feet in altitude."

In 1966 Kay finally secured his bachelor's degree from the University
of Colorado, a double major in mathematics and molecular biology.
The only doctoral program he could find to fit his exacting specification
was the one Dave Evans had established at Utah with a $5 million grant
from Bob Taylor at ARPA. To his own amazement he got accepted
as
the seventh graduate student in the school's tiny department of com­puter science.

"I discovered later that Evans never looked at my grades," Kay said.
"He didn't believe in it. You had to send him a resume, which was all
he ever looked at. He was like Al Davis of the Oakland Raiders; his
theory was to let everybody into training camp and give them a really
decent chance, then be incredibly savage cutting the roster. I was com­pletely thrilled that this guy seemed to think so much of my abilities.
One thing I resolved was that he'd never find out the truth."

Taylor's ARPA money had turned Utah into a hotbed of computer
graphics. Kay discovered that the day he walked into Evans's office to
meet his new mentor. Evans, an introverted gentleman of few words,
reached over to a foot-high stack of documents bound in brown
paper piled on his desk. He handed one to Kay and said, "Take this
and read it."

The title read, "Sketchpad: A Man-Machine Graphical Communica­tions System." The 1963 MIT doctoral thesis of Ivan Sutherland, Taylor's
predecessor at IPTO, the paper described a program that had become
the cornerstone of the young science of interactive computer graphics.
Sketchpad worked on only one machine in the world, Wes Clark's TX-2
at Lincoln Lab. But its precepts were infinitely applicable to a whole
range of increasingly nimble and powerful computers then coming into
existence.
Sketchpad was also, by
Evans's
mandate, the cardinal intro­duction to computing in his doctoral program. "Basically,"
Kay
said, "you
had to understand that before you
were
a real person at
Utah."

Sutherland's system could create graphic objects of dazzling complex­ity, all the more amazing given the severe limitations of the contempo­rary hardware. With Sketchpad the user could skew straight lines into
curves ("rubber-banding"), make engineering-precise lines and angles
(the system straightened out the
draftsman's
rough sketches), and zoom
the
display resolution in and out.
The program
pioneered the "virtual
desktop," in which the user sketched
on the
visible portion of a theoret­ical drawing space about one-third of
a mile
square
(the
invisible por­tions were held in the computer's
memory
and could be scrolled into
view). Contemplating the power of
Sketchpad
was "like seeing
a
glimpse
of heaven,"
Kay
said later. "It had all of
the
kinds of things that the com­puter seemed to promise. You could
think of
it as a light that was sort of
showing us the way."

That graphics could be a directly manipulable
—
and minutely per­sonalized
—
element of the computer interface was one of dozens of
new concepts that bombarded Kay in his first few weeks
at
Utah.
His
mind
on fire, he spent hours in the library stacks photocopying every­thing that grabbed his interest in the computing literature.
He
emerged with hundreds of articles, virtually a living history of comput­ing for his parched intellect to absorb.

He
soon came under other powerful influences.
At
one conference he
heard the oracular Marvin Minsky speak.
Minsky
was an
MIT
psycholo­gist and a computing pioneer, a disciple of the child psychologist
Jean
Piaget
and a founder of the new science of artificial intelligence, which
aimed to reproduce human psychology in the computer.
His
speech was
a "terrific diatribe" about how traditional education destroys the learn­ing aptitude of children, a subject that must have resounded to
the
pre­cocious
Kay's
very soul. Minsky did not specifically prescribe computers
as the answer. But he made intriguing mention of the work a colleague
had done in designing a computer language to help children learn pro­gramming.

Early the next year Kay got to meet this colleague. Seymour Papert was
a burly, bushy-bearded South African, a Cambridge-trained mathemati­cian who managed to combine a single-minded absorption with the
learning skills of children with a profound absent-mindedness about
everything else. Papert had devised a simple programming language
known as "LOGO," the aim of which was to teach children about computers by giving them a tool to see the machine instantaneously respond
to their commands. LOGO literally turned the computer into a toy. Its
most conspicuous feature was a turtle-shaped robot the size of a dinner
plate. This device would crawl about on a schoolroom floor according to
simple commands children could type onto a computer screen: "forward
100" directed it in a straight line 100 turtle steps, "right 90" dictated a 90-
degree right turn, and so on. A pen protruding from the turtles belly
would trace its path on the floor, allowing the more adept of its young
programmers to create patterns of almost limitless intricacy.

LOGO
'S
genius was its ability to turn the abstract (one can command a
computer to do something) into the concrete (one can direct the turtle to
draw a parallelogram). To Kay it was a revelation to watch Papert's ten-,
eleven-, and twelve-year-old subjects use a simple computer to create
designs one would odierwise assume could only be achieved by main­frame systems loaded with complex algorithms. Papert showed the way
toward reducing the machine from demigod to tool (in Wes Clark's
phrase) by subjecting it to the unforgiving scrutiny of children. Kay never
forgot the lesson. As he wrote later, "The best outputs that time-sharing
can provide are crude green-tinted line drawings and square-wave musi­cal tones. Children, however, are used to finger paints, color television
and stereophonic records, and they usually find the things that can be
accomplished with a low-capacity time-sharing system insufficiently
stimulating to maintain their interest." Or as Kay and his colleague Adele
Goldberg wrote later: "If 'the medium is the message,' then the message
of low-bandwidth time-sharing is 'blah.'"

When his turn came to design a programming language at PARC, he
would invest it with several unmistakable elements of Papert's system:
its visual feedback, its accessibility to novices, and its orientation to the
wonder and creativity of childhood. Partially in deference
to
this
last
factor,
he would call it "Smalltalk."