Computing
power was still a relatively scarce commodity in 1974, when
Harold Brown was in his fifth year as Caltech president. One
of the many visitors to Caltech that year was Robert Cannon,
U.S. Assistant Secretary of Transportation for Systems Development
and Technology. "Harold was Secretary of the Air Force when
I was chief scientist of the Air Force" in 1966-1968, says
Cannon, currently the Charles Lee Powell Professor of Aeronautics
and Astronautics at Stanford University. "Harold invited me
out to talk about an urban mass transit program," a $10 million
rail project JPL was considering building. Cannon pointed
out that JPL did not know how to build a railroad, and that
the true cost was much higher than what the government was
offering to pay. Boeing ended up taking the $10 million, and
a huge loss when the true cost turned out to be $98 million.
Communications were further opened up when Cannon, Brown and
JPL director William Pickering had dinner together.
Surmising
that Cannon's tour of duty in Washington was about up, Brown
called his former chief scientist several weeks later and
- speaking on behalf of a Caltech search committee - said
"'Caltech was in need of a boost of strength in engineering,'"
says Cannon. "I was being sought by quite a few people while
I was in Washington. I decided not to be a president of a
university, but I wanted to be in the middle of a top-notch
university operation." Having taught at MIT and Stanford,
where he was also founder of the Guidance and Control Laboratory,
Cannon was attracted to the Caltech experience, and liked
the idea of serving as Chairman of the Division of Engineering
& Applied Science and reporting to the Provost and President.
The three layers of management at Caltech were much more appealing
to Cannon than the 5-7 layers of bureaucracy he faced at other
schools.
Once
in Pasadena, Cannon was surprised that "there was no such
thing as a department," but instead informal groups with "natural-born
leaders" working in particular areas such as aeronautics,
environmental engineering, and applied physics. Among the
first things Cannon did was study what was going on and think
about where engineering needed to go in the next couple of
decades. Computer science seemed to be emerging into its own.
After being a branch of applied mathematics in the School
of Humanities at Stanford, computer science voted to move
into the School of Engineering. At the University of California,
Berkeley, Tom Everhart merged electrical engineering with
computer science. "It was clear to me that computer science
was a very important part of engineering," says Cannon, and
at Caltech computer science was happening around Carver Mead
and his students in electrical engineering.
Cannon
was equally focused on getting to know all of the 70 faculty
in the E&AS division. Carver Mead was "a talented guy and
very focused," but "kind of alone" in the applied physics
group, with the word silicon in his lab name on the door because
he wanted to make sure people took silicon for granted as
part of computing. One by one, Cannon visited with and got
to know all the people in the division and what they were
doing.
In
the negotiations leading up to being hired, Cannon and Brown
talked about acquiring additional top faculty for the division.
Once on the job, Cannon kept Brown informed of what was happening
via informal 15-minute weekly meetings. "Harold is an unbelievably
smart leader who astutely does those things that only he can
do," says Cannon, and he leaves all the rest for others. "I
shared with him one early day that computer science is one
of the things that we have to strengthen."
"I
met more often with Provost Bob Christie so that he was never
surprised by anything," says Cannon. "It was something that
I learned in Washington, the first rule of good management."
One day Cannon informed Christie that he wanted to hire eight
new faculty members as part of his plan to strengthen the
division. Christie was more cautious, and suggested hiring
two or three faculty, as eight was a pretty big jump for a
division whose numbers had been relatively static in recent
years.
"If
you're sitting in Washington as I was before going there (Caltech),
what you do to the degree Congress lets you do it, is you
find and get the best people," says Cannon. So Cannon wrote
Harold Brown, asking him to recall the negotiations leading
up to Cannon's hiring, specifically the part about hiring
more faculty. Part of the plan was that Cannon was to be free
to bring in a good number of new people, "providing I can
see upcoming billets and find ways to afford them." This was
easily done at the time with government funding, particularly
the Defense Advanced Research Projects Agency (DARPA), which
continues to be a major source of computer science funding
at Caltech. Harold Brown sent Cannon's note back by return
mail and in pen wrote on it, "'This is exactly the way I remember
it,'" says Cannon, who shared this will Bob Christie so that
they could move forward together. (He also put Brown's annotated
note in his safe.)
As
a result, Cannon estimates he set up eight search committees
which led to the hiring of 18 new faculty, including 3 in
computer science, during a 5 year period, greatly benefiting
all the groups in the division. It was a marvelous time, when
the whole division shared a cooperative spirit, and all were
looking for really good young people. It was a climate in
which Cannon could bring together the division's faculty to
better get to know one another. But before Cannon moved forward
with a computer science group, he wanted to make sure the
faculty and the trustees, as well as the administration, felt
good about the idea.
"First
of all, I knew Harold Brown would be behind whatever I did,"
says Cannon. "I believed that the trustees were very important,
so I started with the trustees, getting them behind what I
was doing." Rube Mettler, who was CEO of TRW at the time and
is now in the Air Force Space and Missile Pioneers Hall of
Fame, was the key trustee Cannon drove to see. "I went to
see him in his very fine office on top of a tall building
north of Century Blvd," and "he was very cordial," says Cannon.
"I just focused on this one thing, that Caltech being right
at the forefront of technical universities, be at the leading
edge in computer science. His response instantly was, 'It
is unthinkable for Caltech not to do that.'" Thus, it was
set in motion that the trustees would be supportive of computer
science when the time came.
Cannon
organized a retreat for the faculty, because he found that
they knew the people within their group, but not the people
in other groups. A retreat was a way for the faculty in the
different groups to interact and get to know each other and
each others' work. "It was not about computer science per
se. However, computer science was an important new thing out
there," says Cannon, "and they would all be drawing heavily
on computer science, even if they didn't know it yet," whether
they were in fluid mechanics or structures or applied physics
with silicon.
Fortunately,
the infrastructure for a retreat had recently been put in
place. An alert Caltech staff member in Fallbrook, about a
hundred miles south of the campus in northern San Diego County,
saw a "for sale" sign on a 15-acre citrus and avocado ranch
with a 5-bedroom modern ranch house, beautifully landscaped
gardens, heated swimming pool and large underground bomb shelter.
The asking price was $250,000. "What a great place that would
be for a Caltech retreat," he exclaimed when he realized that
the seller was a Caltech alum, a B.S. in chemistry in 1918,
the great film director Frank Capra. Just before Christmas
in 1971, word reached one of Caltech's great benefactors,
fellow chemistry major and Trustee Arnold Beckman, whose charming
and well-chosen words put Capra in the gift-giving mood. By
1972, Frank Capra and his wife Lucille had donated their "secluded,
beautifully situated retreat" to Caltech.
If
ever there was a place for the faculty to forget their everyday
concerns, relax and unwind, it was behind the gates of the
Frank Capra Ranch surrounded by lemon, avocado and orange
orchards. Trying to make time to meet during a hectic week
on campus, forget it, just too much to do, students, research,
etc. But the Capra ranch was simple and rustic, with no ringing
phones or other distractions behind the barbed wire and multiflora
rose fences. It was little wonder that Frank Capra retreated
here for the winter when he wasn't on location making films.
The two hour drive from Pasadena was do-able after work on
a Friday afternoon or Saturday morning, and the winding road
was pleasant enough, with its scenic rolling hills and citrus
and avocado orchards adding a pastoral charm to the chaparral.
Cannon
was planning the perfect weekend. He would invite representatives
from each of the nine EAS groups, perhaps 20-30 people in
all. Any more than that would push the comfort limits of the
ranch house, whose bomb shelter sported a stove, sink, bath,
half a dozen bunks hung ship-style on the walls and a dozen
or so cots. Both the living room and large basement play-room
had fireplaces, and comfortably accommodated 25-30 faculty
members. Hans Leipmann and Frank Marble took time out from
their work in the Guggenheim Laboratory to come. Herb Keller
and the opinionated Brit, Gerald Whitman, came from applied
mathematics. The earthquake group sent its reps (led by Paul
Jennings), as did applied mechanics and each of the other
EAS Division groups.
Under
Cannon, the groups in the division were functioning together
more harmoniously than they had in a long time, and Harold
Brown was proud. "We spent our hours together at the Capra
ranch going around to each group, because it was a meeting
of the whole division, and one person from each group made
a presentation, followed by lots of Q and A," says Cannon.
"It was a chance for the groups to get to know each other,"
and "among other things, the idea of that kind of new department
(computer science) evolved very nicely in the context of the
whole division." The actual date of the retreat is a bit fuzzy,
perhaps because it felt like a vacation and there was no reason
to keep track of time or dates. But it was sometime in 1974
or 1975, and there was enough chill in the air to keep the
logs burning in the fireplace.
Carver
Mead clearly enjoyed throwing three-foot-long logs of well-seasoned
(at least a year old) sawed firewood into the Capra ranch
fireplace, and he kept the fire roaring all weekend. Without
central heating everyone comes together around a fire, a primitive
instinct that had the faculty huddling close together and
bonding while they shared sparks of excitement from Carver
Mead's applied physics presentation, which radically re-conceptualized
computation as a physical process governed by the laws of
physics. Even the applied mathematicians, who had tended to
view computing in terms of mathematical reasoning and logical
operations, became intrigued. "They viewed the computer itself
as uninteresting, just a Turing Machine," says Mead. "So we
talked about how in the world out there computer science had
come to be a mathematical exercise."
"It
is very Caltech to try to understand things down to a fundamental
level where you can make fundamental predictions," says Mead.
On the basis of the underlying physics, Mead and graduate
student Bruce Hoeneisen (who later went to South America to
create a Chilean "Caltech") made some scaling predictions,
which were submitted to the journal "Solid-State Electronics"
in 1971. They argued that in 30 years (by 2001) transistor
size could shrink by a factor of 100, hundreds of millions
of components would fit onto single chips and channel lengths
would shrink to the electron beam masking realm of 0.15 microns
(about where they are in 2001). "People considered it crazy
at the time, and had all these reasons it wouldn't work,"
says Mead. "But it has driven the entire industry," and Moore's
Law derives from scaling.
Mead
could have stopped there, but he was also devising a design
solution that anticipated the implications of scaling (to
smaller sizes with millions of components on a chip) on integrated
circuit and microprocessor (invented at Intel in 1971) design.
Structured Very Large Scale Integrated (VLSI) system design
enabled designers to manage the complexity of systems having
hundreds of millions of tiny transistors, devices and circuits
crammed onto small chips in exponentially more complex arrangements.
"I had been doing VLSI, and it became obvious to me that a
big exponential increase would," says Mead, lead to "increased
distributed computing, decreased centralization, increased
applications, communications and architecture issues...and
not one person is looking at it that way, so here is an opportunity
for Caltech to be at the front of the pack."
The
Division faculty, in contrast to previous decades, came away
from the Capra ranch retreat convinced that a computing science
group was a good idea. "It was really the first time engineering
was proactive about that," says Mead. "So, basically the outcome
of that retreat was that everyone agreed we should try to
find a way to put our good position in VLSI together with
their computer science in a broader way and get something
started."
Cannon
promptly appointed a search committee in the field of computer
science, and the committee's first selection was Ivan Sutherland,
who then took part in the next two selections: Jim Kajiya
and Charles Minter. Sutherland, at 38, was the most senior
of the 15 new appointments to E&AS (all but three of the 15
were in their early or mid-20s). He had been a co-founder
of Evans & Sutherland, and his superb technical prowess,
leading the exciting new arena of virtual imaging, was a perfect
compliment to Mead's in silicon magic. The two - Sutherland
and Mead - were an awesome leadership team in computer science
at Caltech. Awesome.