REGIONAL MEETING: SEEDS OF INNOVATION Drs

REGIONAL MEETING: SEEDS OF INNOVATION
Drs. Alan Kay and Len Kleinrock, TTI/Vanguard Advisory Board
New York, New York · May 13, 2014
“One physicist can keep a lot of engineers busy.”—Alan Kay
Big ideas, and the ability to flesh them out to become inventions with world-changing potential, are the
product of unconventional minds—minds that are simultaneously inner-directed and focused on novel
ideas, rather than looking to the world to reinforce preconceptions. TTI/Vanguard Advisory Board member
Alan Kay estimates that a scant one percent of people alive at any point possess both of these individually
rare personality traits. There are more such people today than 40–50 years ago, yet the production of
transformative invention in the realm of computer science has slid precipitously from its heyday. Both Kay
and TTI/Vanguard Advisory Board member Len Kleinrock are among such thinkers, and both enjoyed the
added good fortune of coming of age intellectually when the funding environment actively fostered the
development of basic science and a long-term outlook through a collaborative meritocracy composed of
brilliant people who came together as a community of researchers. Kleinrock’s contributions were
instrumental in packet switching and the ARPAnet; Kay’s likewise for object-oriented programming and
revolutionizing human–computer interface design with overlapping windows and the Dynabook, with
ARPA funding for his graduate research and with the support of Xerox PARC under the leadership of Bob
Taylor, who headed ARPA’s Information Processing and Techniques Office (IPTO) as its third so-called
director. Perhaps the proper title for the early ARPA/IPTO leaders—J.C.R Licklider, Ivan Sutherland,
Taylor, Larry Roberts, and Licklider, once again—was guide, but certainly not director nor manager. Each
took it upon himself to fund people, not projects, and rely on the people to grow this thing that would
become computer science. Kay, if he were to bestow an award for the IPTO-era advancements, would
give it to the funders, so great and foresightful was their influence. Today, despite an unchanged volume
of relative talent—there still exists that small pool of independent deep thinkers—the research
environment is fundamentally altered such that the inventiveness of the past is largely absent. The need to
chase short-term federal grant money or corporate project funds has fundamentally shifted the scope of
research. The quest for easily explained and quickly picked low-hanging fruit pervades the lives of today’s
graduate students and has become an ingrained part of their training. Both Kay and Kleinrock are on
separate but parallel quests to document an understanding of what released the brilliance of the best
researchers of their era, with a hope of replicating the necessary and sufficient elements to draw the latent
inventiveness out of today’s great young minds. Kleinrock has recently taken on a DARPA grant to collect
oral histories from seminal computer science researchers, with a focus on capturing the essence of what
the environment offered to enable greatness; Kay is in the process of establishing a modern-day, yet
PARC-like, corporate research lab; and both have given deep thought to what needs to be
environmentally present for great research ideas to blossom.
Alan Kay finds it ironic that this discussion on creative greatness should take place in an institution with a
dress code. That the Metropolitan Club requires gentlemen to wear a jacket and tie hardly guarantees
that all who pass through its doors demonstrate elegance. The Metropolitan Club is, he notes,
erroneously conflating rules and principles, goals and vision. Elegance might be the vision, but Kay fit
within the rules by wearing a polo shirt beneath his jacket, sneakers on his feet, and a bowtie snarkily
constructed to look like a $100 bill beneath his chin. Purposefully nonelegant, by his own admission.
Rules get in the way, believes Kay, both in terms of advancing the Metropolitan Club’s goal of being a
classy establishment and in terms of computer science’s quest for intellectual creativity. The research
community—corporate, academic, and governmental—has come to commit its own semantic offense by
conflating invention with innovation. Following Kay’s lead, invention will be used here to signify great work
at the level of basic principles and science (funding category 6.1), while innovation will refer to applied
research (6.2), and engineering will connote technological development and productization (6.3). All three
matter greatly to technological progress, but a single invention can spur a multitude of innovations and
engineering outcomes. Inventions, however, take time. Today’s dominant corporations—Kay singles out
Google and Amazon—are doing phenomenal innovation and engineering, but they are not inventing. In
fact, most of today’s great innovations are traceable to inventions of the ARPA/PARC era. Apple’s iPhone
and subsequent iPad, reminds Kay as a case in point, are instantiations of his 1968 Dynabook; and
Licklider and Taylor sketched out the basics of Siri in the same Apollo-era year.
As Judy Estrin proposed in her 2008 book Closing the Innovation Gap, perhaps the natural ebb and flow
of the intellectual tide features periods rich in scientific invention followed by periods of productive gain
from those gems in the form of engineering-backed innovation. Steven Cherry reminds that this is the
pattern that the industrial-era transportation sector displayed with the invention of the steam engine
preceding by decades its transformative use in the steamboat and railroad. Might such cycles simply be
playing out in computer science? And, if so, will this pattern suffice? Kleinrock believes not: Big problems
for society loom—problems that merit new concepts in computer science. Kay also is not sanguine about
settling with any single form of thinking and problem solving. Society should not be restricted, for it is
through diversity that progress comes.
Yet, diversity does not come of its own accord. Research requires funding, and funding dollars arrive
when granting agencies and corporate decision-makers accept proposals; that is how science proceeds.
The criteria for acceptance are, therefore paramount in determining the ideas that see the light of day. As
noted, Licklider and Taylor both saw the merit of investing in specific researchers, rather than projects,
and then setting them free with an ample timeline and adequate albeit not a torrential outpouring of funds.
This is the model that Vannevar Bush laid out in a report requested by President Franklin Roosevelt, when
World War II was still raging, to develop the science that could raise the country to new heights, and
these were the principles on which ARPA, and specifically the IPTO, was eventually formed in 1963.
Kleinrock and those telling their stories in his oral-history project all repeat the same mantra: The ARPA
leadership—people who were scientists in their own right (note that both Licklider and Taylor were
psychologists, not engineers), purposefully drew to themselves inspirational researchers as program
managers, people with a strong knowledge of the researchers best suited to take each project to the next
level. The leadership delegated authority to the researchers, granting them not only funds but—of equal
importance—the freedom and latitude to be creative. ARPA researchers were expected to debate openly
and feed off of one another’s ideas, all with a collaborative spirit of enthusiastic creation. The culture of
the time was one of equals: A good idea merited attention, whether initiated by a graduate student or a
professor. This was in the days before computer science departments would spring up on university
campuses like dandelions in May. Graduate students and professors alike had backgrounds that ranged
from electrical engineering to physics, and math to psychology and beyond; common among them,
however, was deep study in “really hard fields,” recalls Kay. He and Kleinrock remember everyone
involved working toward a shared vision, but from different angles, relieving the threat of unhealthy
competition while producing an outpouring of great things. The expectation was that seven years was
quite a reasonable span for an idea to mature.
Contrast that bygone research environment to today’s. Now computer science programs have created, if
not a monoculture, then at least a common framework for thought—a framework that produces a jobready workforce, but not one that generates rogue thought and enthusiastic intellectual excursions into
the unknown. Combine an established course of study with the short-termism that pervades the business
climate and has leaked into the governmental funding mindset, and there exist three strikes against
inventive research: no time, no self-directed impetus, and no multiplicity of intellectual discussion or
percolation to stimulate truly new ideas that bubble up to become recognized as such. The role of
graduate students has devolved into one of churning out papers for their professors, rather than doing
creative research on problems that capture their imagination. Kleinrock bemoans not only the loss for the
current generation of graduate students, but also for those they will train up as their own future group
members once they rise to fill faculty slots; after all, it is only natural to perpetuate one’s own experience
as normal and good, even if unknowingly intellectually barren. Arriving from an orthogonal direction—the
U.S. Congress—came an additional arrow in the heart of computer science research that struck in the
form of the 1973 Mansfield Amendment to the then-existing Military Authorization Act, which forbid the
granting of Department of Defense funds—including ARPA’s—to any project that lacked direct military
applicability. This law could well be the singular stroke that pivoted ARPA from being in the business of
having researchers devise their own problems to ARPA itself defining the research problems for which
researchers would than vie to take on. Licklider, who had then regained the helm of IPTO, felt this wound
deeply but had no legal tools to heal it. It was not that Licklider had lost his vision, but rather that he was
forcibly constrained from carrying it out in the post-Mansfield Amendment period. As time has passed,
the perversion of the original vision has only deepened, with today’s program managers too often
confusing “being responsible for being in control,” says Kay. That is, program managers, who now tend
to be bureaucrats, rather than scientists, exert detrimental control. Kay condemns this shift in idea
origination as “defining a dress code for minds”: No tie, no jacket—no moolah. Or, as Kleinrock states, it
is as if ARPA had requested proposals from researchers prepared to deliver 2 bits-per-second to the
foxhole, defined along a tight timeline, rather than the much more loosely defined request for researchers
ready to take on the task of building an Internet. If anything, the intellectual culture of defense funding has
deteriorated further in the interim. The Pentagon has funds available, but only for short stints; internal
oversight involves monthly review of existing projects, making it impossible to consider anything requiring
time-consuming deep thought and exploration.
Challenges to inventiveness come not only top–down, but permeate society more generally. Kleinrock
sees the availability of computing as a notable culprit in its own right. It has become an ever-present
crutch that invites shallow thinking. Today, students in the sciences, including computer science, build a
model, do a simulation, and present an answer without doing important work of understanding what it all
means, or—alarmingly often—even considering whether a numerical result is of a proper order of
magnitude. Questioning why comes as an afterthought, if it arises at all. In some quarters, brainstorming
substitutes for creativity, but the sheen of a quick idea often fades with time. Kay endorses writing down
ideas as they bubble up, but then setting them aside; if the same idea continues to resurface, then
perhaps it is worth pursuing. Kay notes that highly creative people are, by definition, full of ideas; most
don’t go anywhere, but those that stand the test of time can yield greatness. Kay and Kleinrock do not
have the market cornered on ideas for fostering the next generation of basic computer scientists or, more
generally, deep thinkers in any field—or across many fields. Other participants in this discussion on the
Seeds of Innovation offer contributions of their own. One extends Kleinrock’s worry about pervasive
computing to children’s game play, noting that kids increasingly interact with their screens rather than
with one another and their own imaginations devoid of the constraints of a preordained game
environment. “Kids aren’t doing free play anymore,” says this individual, suggesting that tomorrow’s
scientists are losing the capacity for self-driven creative thought and action. This is not to say that the
broad benefits that computing confers should, as a matter of principle, be shelved. One participant
singles out the sheer utility of Wikipedia as an enabler, noting that it is a quick-and-easy means to fill in
voids in knowledge and serve as a jumping off point for further inquiry. Moreover, computing platforms
and networks create opportunities that would be impossible otherwise. For instance, suggests a
participant, the block chain protocol that underpins the BitCoin spec could disintermediate entire
industries if it catches on.
Beyond these, Kay offers the reminder that new horizons remain. The problems of the 1960s and 1970s
retain much yet to solve, and new unexplored problems await identification. What, then, is needed to
transform today’s ecology of inhibited ecology of risk, reward, and potential into one with the vibrancy of
yesteryear to capture the inventive spirit of today’s smartest scientists and technologists? To put this into
context, Kay recalls the terminology that encapsulated the grand scope that Licklider had for the Internet
from its onset. He dubbed it that “intergalactic network,” recognizing that the Internet would eventually be
trimmed down to being merely “galactic,” and that would be good enough. Licklider—and the
researchers he set free with ARPA funding—also recognized that big thinking unavoidably involves failures
and setbacks, but these are best considered, in a literal sense, a cost of doing business. Costs, of
course, matter, yet even during PARC’s heyday Xerox was only investing a fraction of one percent of its
R&D budget into PARC itself. In today’s dollars, this amounted to an annual budget of perhaps $10M.
The payoff? On the order of $1T from the laser printer alone, although Xerox failed to commercialize many
of its other spectacular achievements, which Kay numbers as 9.5 (laser printer, networked personal
computer, bitmap screen, GUI, WYSIWYG and desktop publishing, dynamic object-oriented
programming (via Smalltalk), Ethernet, peer-to-peer and client–server architectures, PostScript, and the
half contribution the invention of the Internet in the form of PUP (PARC Universal Packet), a nonadopted
alternative to TCP/IP. Even for the laser printer, return on investment did not occur overnight, but it did
occur, which is a lesson today’s corporations should take to heart.
Kay uses the language of finance to better connect with business interests he seeks to sway toward
doing invention-level science. The concept of portfolio investment proves a rich analogy. Businesspeople
recognize the potential payoff from investing a portion of resources into very high-risk projects; not every
bet will be a winner, but the gains from the fraction that succeeds more than makes up for the losses
from the remainder. The MIT Media Lab, another go-big-or-go-home research environment, similarly
invites creativity from its researchers. There, the annual budget is $40–50M, which is spread out over the
Lab’s various corporate sponsors and therefore is like a drop in the funding budget for each. Research is
an incredible corporate investment, if done right.
And one corporation is. As previously noted, Kay is currently involved in developing a research lab for an
unnamed corporation of significant stature. Following in the footsteps of Licklider and Taylor, Kay, having
been handed adequate funding for each of two years to recruit the collection of top young researchers he
believes most capable of fulfilling his vision of tackling the big problems in communications, whether
between people and people, people and software, or software and systems; the goal is to “make
communications qualitatively better.”
Beginning one year ago, Kay began setting up this so-called invention lab in an attempt to bring some
properties from the ARPA/PARC model into the present. Past successful properties—such as the use of
visions and considering the type of people to fund—are straightforwardly transportable through time, but
to follow suit with others presents new challenges, notably identifying the best of the talent pool. PARC
had the pick of the proverbial litter by drawing on the ARPA/IPTO community for young PhDs who were
already infused with ARPA vision of interactive computing and pervasive networking.
Kay shares some of his challenges and successes to date: Today there are people just as talented and
productive—and maybe even more so—as in the past, although as the ranks of people pursuing
advanced degrees in computer science, the true stand-outs represent a smaller percentage of recent
graduates. Equally relevant, corporate research centers can attract these individuals with the right kinds of
freedoms, funding, visions, and the ability to share their work with those outside the research center. Kay
notes that PARC faced this same set of obstacles and successfully resolved then; they can, he believes,
also be resolved today. The budgets of ambitious research centers are now—as then—quite low,
because very high-quality researchers are still quite rare. Yet, as PARC demonstrated, even with a spare
headcount, much can be accomplished; recall that PARC did most of its inventions with roughly 25
principal researchers, along with students and a modest support staff. Kay has every reason to expect
reason an organization of a similar size to be every bit as effective today.
A typical burdened headcount for an internal "futures organization" would be on the order of 40 people, or
about $8M–$10M dollars annually. Even the smallest Fortune 500 companies—those with perhaps $5B in
revenues—could easily afford this 0.2% investment; larger organizations—whether corporate or, say,
divisions of the Department of Defense—would find it even easier to devote a few million per year toward
funding an invention lab. Most companies, in fact, would say they have something like this, but a key
component is missing. The key to making this work is to find and tolerate a small yet select group of very
different kinds of highly trained people, rather than trying to pull one together from the already hired highly
trained employees who are adept at meeting the shorter term needs of the company. It is this latter error
in thinking that causes most companies to fail when they try to do establish a lab. Instead, they should
embrace the diversity of abilities needed for invention, just as they do for their other processes: They need
to find those who have “ability to invent fruitful futures.”
References mentioned by speakers and participants:
 http://www.ttivanguard.com/regionalmeetings/dcregionalhighlights.pdf Robert Charette, Alan Kay, and
Len Kleinrock, Innovation in Big Organizations is Hard But Not Impossible, TTI/Vanguard Regional
Meeting, Washington, D.C., May 2014
 Donald Brown (1991): Human Universals.
 http://www.nsf.gov/od/lpa/nsf50/vbush1945.htm — Full text of Bush, Vannevar. 1945. Science: The
Endless Frontier.
 http://u-tx.net/ccritics/as-we-may-think.html — Full text of The Atlantic article: Bush, Vannevar. 1945.
As we may think.
 Judith Estrin (2008): Closing the Innovation Gap.
 http://www.kurzweilai.net/memorandum-for-members-and-affiliates-of-the-intergalactic-computernetwork — Full text of Licklider, J.C.R. 1963. Memorandum for members and affiliates of the
Intergalactic Computer Network.
 http://memex.org/licklider.pdf — Full text of two seminal papers: Licklider, J.C.R. 1960. Man–computer
symbiosis. LIcklider, J.C.R., and R.W. Taylor. 1968. The computer as a communications device.