Rapport sur les orientations de la recherche technologique dans la Silicon Valley. « What’s Left to Know ? »

research in a total-connect worldconversations about Tech Research Futures

3

This is about a set of conversations. Indeed, a conversation that started in

February 2011 as an introspection about the balance between academic and com-

mercial research – where was the important research in IT and telecoms being done?

Who was investing in it? To whom and for what is the long-term R&D world even

relevant anymore?

Those conversations continued and started involving others, including the author

of this remarkable set of interviews. Things got pretty transgressive: what if what we

see before us really is it for the next 20 years? What if most of the important tools and

methods we utilize for the next few decades are the cutting-edge resources of today?

If so, we are just exiting a period of radical transformation, and entering a massively

incremental mode.

To explore where research is going, you have to talk to practitioners. And, you

have to speak in their language, understand their issues, and know the canon. We

were able to find that interlocutor in the person of Lee Gomes. Known to many in Sili-

con Valley for his reporting on behalf of national publications such as the Wall Street

Journal and Forbes, Lee was the perfect entry-point into a wider discussion. The result

is what you have here --- a rich set of discussions with some of the companies and

universities responsible for the future of high-tech advanced research. Those inter-

views plus our own first hand observations and interactions with peers and major

customers formed the basis for this report.

We at Orange share these findings in the spirit of exchange and the idea of innova-

tion as open and public. While what follows is ostensibly about research, we recognize

that it is fundamentally about the Talent that drives innovative research. The concentra-

tion of Talent in Silicon Valley is unique in the world – and highly sought after. Respect

for Talent, recognition that Talent is highly mobile and votes with its feet -- this is what

drives the questions posed here. Is that PhD as important as that startup? Is munging

some ugly public data set as rewarding as working on the Facebook graph? These are

not questions about technology, but about where Talent wants to go and what it wants

to accomplish. Let’s add to that: when it wants to accomplish it.

- Georges Nahon

CEO

-Mark Plakias

VP

Orange Silicon Valley

Editorial: Talking about Talent

5

At Orange Silicon Valley, we are actively engaged in most

of the topics cited in these discussions with researchers in

the IT, communications, and online media and commerce

industries. In the course of this ongoing collaboration the

voices from both industry and academia have spoken and

their message seems clear: life in the “Labs” -- and beyond --

is never going to be the same. Almost daily we can observe

a significant number of impactful innovations manifesting

themselves in commercial products and services that are

the fruits of smart people working outside of the corporate

or academic R&D sphere. Who are these smart people?

They’re not wearing white coats, nor do they always

publish papers. From the corporate R&D establishment’s

perspective, they look like barbarians, or something alien.

The facts of the matter are clear from our perspective:

things are running very fast in a digital and networked world,

and what research does and how it is done probably needs

to be revisited. The title of this report phrases the current

state of affairs in information and tech research as a question

in epistemology, but ‘what’s left to know?’ is also a question

about tempo and scale. We know big corporations spend

most of their time protecting the most profitable part of their

business now, but digital platforms are impacting these

centers so quickly that further R&D in these disrupted and

discarded areas seems somehow suspect.

This is happening everywhere, including the deepest

recesses of corporate R&D, where cycles of investment

and divestment are getting shorter. The deterioration

of corporate trust for research is probably based on

management impatience with too-long research cycles,

squeezed by more and more competitive scenarios with

barbarians coming from all over the place creating new,

hugely successful products from well-funded disruptors such

as eBay, Amazon, Google, Zynga, Facebook, Yahoo, etc.

In this atmosphere, R&D doesn’t seem to be

producing anything competitive with what the barbarians

produce and deliver. A condescending view is that Corporate

R&D projects may be taken by surprise by the agility and

audacity of the barbarians -- who have no faith and no

respect for the laws or rules governing the establishments

(if they ever even knew them), or indeed for any legacy.

For them, risk is the new normal. Risk seems to not carry

the perception of threat, as they have nothing that can be

threatened

But is this just attitude, or is it data-driven? The

idea that analytics and pattern-recognition of very large data

sets are both basic research and a model for how research

gets done is an important topic in these discussions. As

Facebook’s Cameron Marlow puts it in his interview: “The

social interactions on the Internet, and on Facebook in

particular, are at a level of detail and scale that haven’t

existed before. They allow us to answer questions about

social interaction and forces that we’ve never been able to

answer...” Although SRI’s Winarsky does not see Zynga’s

“new type of business proposition” as anything resembling

research, he does describe it as “even more valuable than

technology.” Google’s Peter Norvig sees the availability of

these data sets as unique to industry, and a reason to migrate

...things are running very fast in a digital and networked world, and what research does and how it is done probably needs to be revisited.

What’s left to know?

6

there: “...there are some things you can’t do in universities,

and that’s one reason why I am not in the university now.”

It was not always this way. In the past, game-changing

companies were fewer in number, therefore movements of

the industry were calibrated by other incumbents’ R&D labs

performance in a time-scale equal for all players. More

importantly, everyone was staying in their own business

territory -- until digital changed everything and made every

company look like it was playing in the other guy’s garden.

The tacit rules of mutual control -- “do not come into my

territory and I will not get into yours” -- are over because of

all things going to digital.

What are the larger ecosystem implications of this

shift? For one thing, important, viable research work at the

core of computing and communications seems somehow

not as visible or compelling as it used to be. While this is

a contestable statement, and there are rich discussions

about this point inside these pages, one incontestable fact

is that innovation is redistributed and shared with start ups

-- and some of these startups come from academic projects,

now encouraged by private investors. UC Berkeley’s

David Patterson recognizes this trend, which he views as

manageable: “What’s happening is that our students are more

tempted today with some new and exciting technologies

rather than finishing their PhDs, so they decide to go and start

a company.”

The friction of corporate distrust of what’s happening

internally in R&D is discussed at various points in these

pages as management seeing all this money and time spent

developing new innovative products, but getting beat to

market by barbarians. This seems to be forcing management

to succumb to the temptation to kill years of effort, with the

possible exception of some patents that can be monetized.

In this oft-repeated scenario private R&D seems more and

more like a defensive move (get more and more patents filed

and registered) than an attacking one.

More recently, events such as the HP’s acquisition of

PALM and its coffer of patents , the acquisition of Nortel’s

patent portfolio by a consortium of IT companies and the

Google/Motorola multi-billion dollar patent portfolio grabs

have thrown this strategy into a new light: one where the best

defense is a good offense. In this new competitive scenario,

where intellectual property becomes an offensive weapon,

hundreds of man-years of corporate R&D can be used to

drive licensing claims which add $5 to the cost of a mobile

phone. Whether this creates innovation or just lawyers’

fees is open to debate, and is just one of the many ways

in which the discussion about how innovation and research

interact continues to evolve. It is a moving train. And it may

be that this new model is going to forever complement (and

better) other more traditional forms of R&D that may survive

in the fields of IT, communications, and online media and

commerce. At Orange Silicon Valley our conviction is that

we cannot understand alone what’s inside a moving train

by watching it from the platform. We need to be in the train

ourselves along with others to address the question, “What’s

new to know now?”

The tacit rules of mutual control -- “do not come into my territory

and I will not get into yours” -- are over because of all things

going to digital.

This seems to be forcing management to succumb to the temptation to kill years of effort, with the possible

exception of some patents that can be monetized.

8

Most have heard the story of the woeful idiot who happened to be U.S. director of patents in 1899, and who urged that his office be closed because there was nothing

remaining to invent. In fact, that tale is an urban legend with no basis in fact; Charles H. Duell, who held the post at the time, far from being pessimistic about future discoveries, was actually a booster when it came to what Yankee ingenuity would make possible, akin to “You ain’t seen nothing yet.” (Quotes suggesting the opposite have been proven by scholars to be fabrications). But the fact that someone, somewhere, even bothered to create the historical falsification in the first place can be taken as a warning of the dangers involved when a given generation becomes so engrossed in its own repertoire of massive inventions that it devalues longer-term perspectives. But - and to use another problematic formulation - what if this time, it’s different? What if this time, they’re right, that the moment we’re in dwarfs everything up until now by comparison, and requires our full attention and ingenuity? To be clear, we are not talking here about the totality of science and technology, but instead, a subsection of it, in the enormously important world of IT, especially as it involves data and data sciences. What if the fundamental data-intensive infrastructure of computers - Moore’s Law, new data base tools and

ubiquitous mobile devices - is now in place, the way the basics of the automobile were in place once the modern internal combustion engine was realized? What if developments in IT and data from this point on are all incremental? This, of course, does not imply that there won’t be any major technology improvements, or shifts in corporate fortunes, in coming years and decades. After all, both cars and car companies look drastically different today than they did 50 or even 25 years ago, despite having the same technological foundation. But the changes were essential, gradual and incremental, rather than being earth-shaking and revolutionary, as they were in Henry Ford’s time. To understand why we are even raising the question of the future of data and IT, let’s first consider the classic view of the subject - with a bow to particle physics, we can almost call it the “Standard Model” of Silicon Valley. In this telling of history, most IT research was done at one of two locales: inside academia, under the sponsorship of the U.S. government, and at select groups of large companies with dominant market positions and the healthy profit margins that accompany them. The research-to-product transition followed a traditional path, beginning in a lab somewhere and then making its way to the marketplace, perhaps pausing along the way to accumulate some venture capital funding. How little that looks like today’s world. For one, federal funding for research has diminished or been reallocated.

History has not been kind to those who managed to

become associated with the idea that everything that

can be invented already has been.

by Lee Gomes

9 9

Second, the great standalone corporate research labs of the post-war era, with AT&T’s Bell Labs (now Alcatel-Lucent Bell Labs) being the preeminent example, have been repurposed, if they even exist at all. Perhaps most significantly, technologies that not too many decades ago were but fledgling research ideas have today become robust, even commodity, product categories - notably highly integrated semiconductors, high-capacity storage devices, extremely sophisticated software and robust ubiquitous networks capable of high-capacity digital communications. The result is on display everywhere in Silicon Valley. Dense chips, cheap storage and ubiquitous networking have created the world of “Big Data,” in which hundreds of millions of computers and mobile devices are creating staggering amounts of information. Rather than the clock-like migration of technologies from research labs to companies, what we see instead is the steady ascent of what might be called the “research giants” - the best example, of course, is Google. These companies, arguably, are the only institutions with the resources, both capital and human, to handle data at the levels it is being created. It follows that they have privileged access to the massive data sets that enable the type of analytics and research about human behavior that is creating the wealth of the 21st Century. To be sure, part of

the innovation of these new giants is ecosystem-specific: Google and others share, via open source, the tools used for managing and learning from Big Data, and actively encourage these open source resources adoption by individuals and entrepreneurs. But the new reality is that the best, and most massive, data is in private hands. Far from being content, almost passive, “end-users” of academic research, as they might have been in the past, the big Silicon Valley companies of today are arguably doing most of the heavy lifting in IT research in the first place. In effect, is it perhaps the case that the task has fallen to the Google’s of the world, rather than any traditional “research” operations, of providing whatever incremental improvements remain to be made in the art and science of data? In this document you will hear from thoughtful practitioners on both sides of this question. By contrast, academic computer science departments find themselves doing a severely resource-constrained imitation of the commercial efforts underway at the big Silicon Valley companies. To ameliorate this, collaboration is necessary, and indeed happening: companies such as Google, SAP, Amazon, and Huawei are supporting long-term university research efforts such as University of California Berkeley’s AMP Lab. Still, the big picture seems weighted towards

“

“

Put in its most stark fashion, as far as data-related IT goes,

does traditional research even matter any more? That

question is at the core subject of this report.

10

In one sense, there is nothing particularly shocking about suggesting that parts of the IT industry have become so mature that changes from here out will be incremental. It has happened consistently over the last two centuries, in everything from steam power to electricity to radiography to telephones and televisions. All went from being a newly-discovered phenomenon of nature, the domain of researchers and scientists, to being the basis of large commercial enterprises, the domain of business people - with the occasional assist from engineers. Even in science itself, research can reach a mature phase. Isaac Newton did a fairly complete job of describing the movement of everyday objects in the everyday lives of human beings, from apples to planets. Early in the 20th Century, we discovered that Newton’s Laws didn’t hold for the very small or the very fast. But quantum mechanics and relativity don’t repudiate Newton, but instead modified him

for new domains but leaving unchanged the many technologies built on Newtonian principles. We hope readers appreciate that when we ask “Does IT research still matter?” we are asking it not because we view the issue as settled, but instead, in an attempt to provoke thoughtful discussion. Some readers might be shocked that

anyone would question the value of research. (In this report, “research” means basic, unstructured research, with no obvious short-or mid-term connection with a company’s existing product lines). But in actuality, there has always been a surprising lack of consensus about the economic value of research, at least when it is performed by an individual company. (Few doubt that federal dollars on even the most basic, untested forms of research are well spent). Indeed, one

immediate engagement in a commercial context: these companies are so wealthy that they regularly poach some of academia’s most talented faculty members and students, further depleting academic research efforts. If you are interested in exploring some aspect of the “social graph,” where would you rather be: at even the best-funded academic department, or at Facebook, where you would have access to a 25-petabyte Hadoop cluster? And from which do you think the most interesting insights will spring into how people use “social media?” Put in its most stark fashion, as far as data-related IT goes, does traditional research even matter any more? That question is at the core subject of this report. This idea for this project began when a small group of us, sitting together at a table, made a simple assumption: Everything else about the world has changed on account of the Internet; why should the practice of research

be any exception? We interviewed researchers, research managers, entrepreneurs and other deep thinkers who spend their careers in Silicon Valley. The questions were all designed as variations on the same theme: How should technology companies, in an age of ubiquitous mobile computing, “big data,” shifting business plans, shortened investment horizons, be thinking about “research?” The bulk of this report is devoted to those interviews.

If you are interested in exploring some aspect of the “social graph,” where would you rather be: at even the best-funded academic depart-ment, or at Facebook, where you would have access to a 25-petabyte Hadoop cluster?

many academic economists, who caution against attempting to draw any conclusions at all from publicly-reported accounting data, since the law allows companies considerable latitude in what they report as “research.” One paper by a trio of economists, with more stridency than is usually associated with academic writing, called the report “extremely misleading,” saying that in ignoring decades of prior literature, it “mixes incorrect conclusions obtained from an uninformed and simplistic analysis with some common sense advice” analogous, to telling people, “It is better to be rich and healthy than poor and sick.”1

There are certain questions about research that one doesn’t need to be a trained economist to raise. For one, there is plenty of anecdotal evidence to suggest that large research operations

writer said that as far as companies are concerned, research might best be defined as a “faith-based initiative.” Consider the contradictory conclusions reached on the matter by two different groups of professionals: business school professors and business consultants on one hand, and academic economists on the other. Representative of the former category is a series of annual reports by Booz & Co. that began in 2005 and have continued since. Analysts at the firm say they have been able to find no correlation between R&D spending with just about anything most companies care about, such as sales growth, profits and market cap. (Though the report noted that the bottom 10% of R&D spenders tended to underperform in other areas as well). The report did not sit well with

Experimental modes of transistors, 1953

Bell Labs is an excellent case study in the difficulties of knowing what sort of value to assign to in-house research.

11

12

often offer little protection to their companies in anticipating and responding to new business challenges. Sometimes, this is hardly the fault of researchers, as in the “innovator’s dilemma” situations where management simply can’t bring itself to undergo what might well be the wrenching business model changes necessary to adjust to shifting technologies. But what do we make of Microsoft’s seeming lack of ability to anticipate or out-maneuver Google, or Google’s current difficulties in keeping up with Facebook?” The “research skeptic” would also note the enormous values attached to companies that seemed to be based on no research at all, but instead on a entrepreneurial insight that was perfectly executed. Facebook, Twitter and Groupon all come to mind. Bell Labs is an excellent case study in the difficulties of knowing what sort of value to assign to in-house research. There is, of course, no doubt that the Labs contributed enormously

to human knowledge during its heyday in the years before the break-up of AT&T, probably more so than any single institution on the planet. (Transistors, evidence of the Big Bang, Unix and C; the laser; quantum computing breakthroughs: the list goes on and on). But how about AT&T shareholders; from their admittedly provincial point of view, did they get their money’s worth? Answering that question is probably impossible, as it involves unraveling a tangle of hypotheticals and counterfactuals, and nearly everyone we asked had a different answer. Most, in fact, expressed agnosticism. The debate about the value of corporate research strongly resembles the debate among economics about the value of “free trade.” Is the “engine” of prosperity, or merely it “handmaiden?” Put differently, does free trade cause fundamental economic growth in the first place, or does it simply accompany it after the fact, like an attendant in a bridal party? The best

1956

There is, of course, no doubt that the Labs contributed enormously

to human knowledge during its heyday in the years before the

break-up of AT&T, probably more so than any single institution on

the planet.

13

characterizations of their research efforts depending on the audience. With Wall Street, they might emphasize its leanness and its tight connection with product groups and quick commercialization. With prospective employees in graduate schools, they are likely to give the impression that new hires are able to pursue their field’s deepest problems - without being distracted by any noise from the grinding wheels of commerce. An additional issue is that many people in Silicon Valley are unaware of the area’s actual history, especially as it involves the relationship between basic research and corporate success. Most people acknowledge obvious well-known facts, like the role that DARPA played in the creation of the Internet. But the pervasiveness of federal involvement in creation of the IT industry is often underappreciated. At nearly every step of the way, federal funding was involved with major Silicon Valley developments. In the 1960s, the first customers of semiconductor products were the military, who displayed ample patience as chip companies worked out the bugs in their earliest efforts at fabbing chips. It was just as true in the 1990s at the creation of Google, since

evidence for the latter hypothesis is that nearly all countries, the U.S. included, take a protectionist approach to their nascent industries. Rephrased to deal with the research issue, the question becomes, “Does research make companies rich, or can companies do research only if they are rich in the first place?” Now that we have advanced our critique of the “Standard Model” of research, and suggested why much of Silicon Valley is perhaps unknowingly in a “post-research” phase in its history, we are obliged to challenge what we ourselves have been arguing. First, we are required to point out is that when trying to figure out a given company’s approach to research, the least reliable source of information is often the company itself. For one, accounting rules about research are so ill-defined that no two companies are reporting the same activities when they report on their “research” budgets. (Note that the authoritative reports by the National Science Foundation about corporate and government R&D spending do not rely on public accounting data, but instead on confidential, anonymized information shared with government by industry). Companies will also adjust their

Does research make companies rich, or can companies do research only if

they are rich in the first place?“ “

14

Larry Page and Sergey Brin did their work on the “Page Rank” algorithm while being funded from a grant from the NSF. Another problem is that “breakthroughs” are almost never as simple as they seem, and rarely occur in isolation. At the same time Brin and Page were doing their work, two other research groups had essentially the same insight involving how a page’s link structure could be mined for crucial information about the page’s reliability. Later, as Google engineers were scrambling to adjust to their company’s astonishing growth, their seeming improvisations occurred inside a considerable ecosystem of existing ideas. One of many examples is the Paxos Algorithm, developed in 1990 by Leslie Lamport, then a DEC researcher, which provides a way of dealing with results provided by potential unreliable computers, a clear problem in a massively parallel data system like the one Google’s engineers were building. (Incidentally, while many fans

of Google think the company emerged fully-formed from the heads of its two founders, the company itself is the first to acknowledge the extent to which its own considerable engineering efforts relied on earlier work by others).The point of all this is that it may be problematic to assume that Silicon Valley operated one way in the last era and an entirely different way today. Descriptions of each era are highly anecdotal; for every example of one pattern for the path that research might take going from lab benches to store shelves, it’s possible to come up with a competing narrative. Consider storage. Much of the pioneering work on the modern magnetic disk drive was done at IBM during the 1950s, and companies have been responsible for the most of the subsequent research responsible for the continuing increases in areal density - the storage equivalent of Moore’s Law. (The cost of storing a unit of information on a disk drive is now 122 million times cheaper than it was in the 1950s). But dramatic disk drive innovations have also occurred in academia, such as the breakthrough notion of RAID storage, developed at UC Berkeley in the 1980s as a way to get highly fault-tolerant storage even while using low-cost, commodity disk drives.

Another critique of our data-oriented hypothesis is that it might unknowingly be a symptom of what has been described as the tech world’s current data fetishism. In many areas, such as language translation and speech recognition, the massive amounts of data available today are allowing companies to fully implement the statistical “machine learning” techniques developed in the 1980s, following the failure of traditional “rule-based” AI. Google Translate is the preeminent example here; while far

it may be problematic to assume that Silicon Valley operated one way in the last era and an entirely different way today.

“ “

15

& observation that academic computer science has been “overtaken” by data science that can be done better by the big tech companies; that academic computer science departments seem to be lagging, not leading, in innovative ideas. One explanation of this apparent phenomenon involves our hypothesis that all the basic work has already been done, which is why academic research seems to so closely resemble commercial research. But another interpretation is that universities are faced with a decline in federal research dollars, and so to attract corporate sponsors, they must essentially pander to what they assume to be the current preoccupation of potential funders. We must also be careful about assuming that the yawning gap that now exists between a Google and virtually any other company or academic department will remain forever. In fact, several forces are at work to narrow it. Improvements in disk storage continue apace; a petabyte 3.5 inch drive should cost $250 or so within the decade. In addition, there are many efforts underway to turn the building and running of a giant data center into a commodity undertaking, no more difficult than setting up a corporate LAN. These efforts are occurring in academia as well as in a new breed of Silicon Valley startups dedicated,

field (much of which will no doubt occur in the academy). Further, it’s possible that we will discover that as long as human beings are involved, past results are no indicator of future performance, no matter how much data one has. All of which are reasons that many people view with alarm any attempt to minimize the importance of research. Companies, this school of thought holds, have a natural tendency to be “free-riders;” to not want to pay for things that don’t obviously and immediately benefit them. Basic research, which even its most traditional supporters say is unpredictable in its distribution of benefits, is precisely the sort of thing that companies in the current investor climate are likely to avoid. Those concerned about the overall state of research today worry that at the very moment that market pressures are forcing firms to pull back on R&D, the government is under economic pressure of its own, and is not able to play its familiar role and take up the slack. Thus, the basic view is that by ignoring research, we won’t be creating the ecosystem that will allow the next Google to be formed. In a similar vein, these people would have a different interpretation of current events than those offered by critics of the Standard Model. For example, it’s common now to hear the

from perfect, its ability to allow a basic understanding of almost any text in any language is astonishing. But the current enthusiasm for data goes beyond implementing machine learning algorithms. Especially in e-commerce, it is assumed that the vast amounts of data we leave behind via our mobile phones and computers have some secret key to our future behavior. The middling success that companies like Amazon and Netflix have with their recommendation systems suggest that much work remains to be done in this

Those concerned about the overall state of research today worry that at the very moment that market pressures are forcing firms to

pull back on R&D, the government is under economic pressure of its own, and is not able to play its familiar role and take up the slack.

16

for example, to providing versions of Hadoop and its related tools that an average IT shop can use. It need be noted that even we, in our deliberately provocative role questioning assumptions about research, would have to agree that there are potentially revolutionary breakthroughs on the horizon. An obvious one involves quantum computing. Computer scientists disagree about the repertoire of problems that a quantum computer could effectively take on. But in the very least, they will force us to rethink the encryption systems currently responsible for all Web commerce. Another potential breakthrough involves a fundamental algorithmic advance in learning how to parallelize computing problems. Microprocessor companies have long since given up on making a single chip that runs ever-faster; chips today ship with four or eight (or even 64) “cores,” each of equal power. But just as it takes nine months to make a baby, most software problems need to be solved in order, one step at a time. A fundamental algorithm to change that - not that anyone has any idea of what it might look like - is inevitable . Finally, we’d also like to point out that even in a world of constant change, some things endure. One is the role played by venture capitalists. VCs have never seen themselves in the business of funding basic research, and certainly don’t these days. Outside of biotechnology - an entirely different

undertaking, where the investment calculus is entirely different - VCs say that their ideal remains research whose commercial appeal is demonstratively obvious. Or, as a spokesman for Sequoia Capital, currently one of the most successful of the fabled venture firms along Palo Alto’s Sand Hill Road, “In our little corner of the world, we get involved in the “D” part of R&D. We leave it to the very creative and very capable talents at universities, government labs and corporate centers to dream up a world of new possibilities.” In closing, there is no doubt that the massive scalability of Internet-based businesses has changed the way we think about research. The urgency created by these scale effects is based on the sheer amount of data available: which poses not just monetization but research opportunities that are here and now. The world has changed, and in the following section we summarize the contrasts in that journey from then to now.

We must also be careful about assuming that the yawning gap

that now exists between a Google and virtually any other company or

academic department will remain forever. In fact, several forces are at

work to narrow it. Improvements in disk storage continue apace;

a petabyte 3.5 inch drive should cost $250 or so within the decade.

“

“

Soft

war

e

Serv

ices

Har

dwar

e

Org

aniz

atio

ns

1985

1960

1995

2010

1940

1990

2005

1980

2000

UN

IX o

pera

ting

syst

em

deve

lope

d at

AT&

T’s

Bel

l La

bs, o

fferin

g m

ulti-

user

, m

ulti-

task

ing

feat

ures

. To

day’

s M

ac O

S X

is a

de

scen

dent

.

Dig

ital E

quip

men

t Cor

p in

itiat

es d

evel

opm

ent o

f a

new

ope

ratin

g sy

stem

, co

de-n

amed

Sta

rlet,

whi

ch w

ill be

com

e th

e w

idel

y-us

ed m

ulti-

user

op

erat

ing

syst

em k

now

n as

VM

S.

Insi

gnia

Sol

utio

ns

intr

oduc

es S

oftP

C

Win

dow

s em

ulat

or

for

Sun

wor

ksta

tions

, us

herin

g in

des

ktop

vi

rtua

lizat

ion.

Lesl

ie L

ampo

rt a

t DE

C

deve

lops

the

Pax

os

algo

rithm

, lat

er u

sed

by

Goo

gle.

LIN

UX

oper

atin

g sy

stem

de

vcel

oped

by

Linu

s To

rval

d, b

uilt

on o

pen

softw

are

licen

sing

By

June

of 2

010,

LIN

UX

was

the

OS

for

the

10

fast

est s

uper

com

pute

rs

on th

e pl

anet

.

VM

War

e is

foun

ded

and

pate

nts

its h

yper

viso

r vi

rtua

lizat

ion

tool

s.

Goo

gle

intr

oduc

es a

so

ftwar

e fra

mew

ork

for

man

agin

g la

rge

data

se

ts a

cros

s di

strib

uted

co

mpu

ter

cent

ers

calle

d M

ap R

educ

e.

Yaho

o! a

nnou

nces

it

has

laun

ched

the

larg

est

inst

ance

of H

adoo

p da

ta m

anag

emne

nt

softw

are

runn

ing

in a

cl

uste

r of

10,

000

Linu

x pr

oces

sor

core

s.

Wor

k be

gins

on

UnQ

L,

Uns

truc

ture

d Q

uery

La

ngua

ge, r

esul

t of

NoS

QL

mov

emen

t.

John

Bar

deen

and

W

alte

r B

ratt

ain,

Bel

l La

bs in

vent

s tr

ansi

stor

Gor

don

Moo

re

publ

ishe

s ar

ticle

in

Ele

ctro

nics

cla

imin

g de

nsity

of c

ompo

nent

s on

an

inte

grat

ed c

ircui

ts

doub

les

annu

ally.

Leon

Chu

a pu

blis

hes

a pa

per

desc

ribin

g ba

sic

prin

cipl

es o

f a M

emris

tor

Son

y an

d P

hilip

s co

llabo

rate

to

stan

dard

ize

Com

pact

D

isc

audi

o fo

rmat

.

Bel

l Lab

s in

trod

uces

the

first

meg

abit

mem

ory

chip

.

HP

ann

ounc

es

deve

lopm

ent o

f a

Mem

risto

r ba

sed

on

titan

ium

dio

xide

film

.

HP

ann

ounc

es

deve

lopm

ent o

f a

Mem

risto

r ba

sed

on

titan

ium

dio

xide

film

.

Ste

war

t Bra

nd a

nd L

arry

B

rillia

nt s

tart

a d

ial-u

p B

BS

cal

led

The

Wel

l, w

hich

bec

omes

the

first

on

line

virt

ual c

omm

unity

an

d on

e of

the

first

co

mm

erci

al d

ial-u

p IS

Ps.

The

wor

ld’s

firs

t web

site

is

cre

ated

at C

ER

N,

the

empl

oyer

of T

im

Ber

ners

-Lee

, cre

dite

d w

ith m

arry

ing

hype

rtex

t to

IP p

roto

cols

.

Firs

t SM

S fr

om

phon

e-to

-pho

ne s

ent i

n Fi

nlan

d.

Sal

esfo

rce.

com

the

wor

ld’s

firs

t sof

twar

e-as

-a-s

ervi

ce (S

aaS

) co

mpa

ny fo

unde

d by

ex

-Ora

cle

exec

s.

W3C

rele

ases

firs

t pu

blic

dra

ft of

its

Web

S

ervi

ces

Arc

hite

ctur

e w

hich

sho

ws

inte

rope

rabl

e so

ftwar

e sy

stem

s co

mm

unic

atin

g an

d de

scrib

ed v

ia X

ML.

Y-C

ombi

nato

r cl

oud

stor

age

star

tup

Dro

pbox

is

foun

ded,

five

yea

rs

lart

er io

s be

ing

valu

ed a

t $5

billi

on.

Mar

k Zu

cker

berg

la

unch

es fa

cebo

k.co

m

web

site

.

App

le’s

App

Sto

re o

pens

in

Jul

y as

an

upda

te to

iT

unes

, iP

hone

3G

shi

ps

the

next

day

with

it

prel

oade

d.

Wes

tern

Ele

ctric

La

bora

torie

s an

d an

en

gine

erin

g de

part

men

t of

the

Am

eric

an

Tele

phon

e &

Tel

egra

ph

com

pany

con

solid

ated

to

form

Bel

l Tel

epho

ne

Labo

rato

ries,

Inc.

Fam

ous

Silic

on V

alle

y ne

twor

king

org

aniz

atio

n C

hruc

hill

Clu

b ho

lds

first

mee

ting,

with

ke

ynot

e by

Rob

ert

Noy

ce, i

nven

tor

of th

e in

tegr

ated

circ

uit.

Vete

ran

entr

epre

neur

P

aul G

raha

m c

o-fo

unds

Y-

Con

mbi

nato

r, w

hich

be

com

es th

e on

e of

the

mos

t pow

erfu

l sta

rtup

in

cuba

tors

of m

oder

n tim

es.

Goo

gle,

Mot

orol

a,

HTC

and

oth

ers

form

O

pen

Mob

ile A

llianc

e to

pro

mot

e A

ndro

id

ecos

ytem

.

IBM

Res

earc

h (T

he

Wat

son

Sci

entifi

c C

ompu

ting

Labo

rato

ry

was

foun

ded

at

Col

umbi

a U

nive

rsity

in

New

Yor

k)

Apa

che

Sof

twar

e Fo

unda

tion

is fo

rmed

to

sup

port

Apa

che

web

ser

ver

softw

are.

SR

I Int

erna

tiona

l fo

unde

d as

a s

epar

ate

rese

arch

inst

itute

from

its

par

ent,

Sta

nfor

d U

nive

rsity

. PA

RC

fo

unde

d as

rese

arch

ar

m o

f Xer

ox C

orp.

The

Res

earc

her’

s Jo

urne

y:

Mile

ston

es in

Tec

hnol

ogy

R&

D

A

BC

D

E

F

AB

CD

FA

BC

DE

FG

H

IA

BC

DE

FG

H

A -

192

5

A -

194

7

B -

194

5

B -

196

5

C -

197

0

C -

197

1D

- 1

980

D -

198

5

E -

1984

F -

2008

G -

201

0

E -

1999

F -

2005

G -

200

7

A -

1969

E -

1991

B -

197

5F

- 19

98C

- 1

988

G -

200

4D

- 1

990

H -

200

8I -

201

1

A -

198

5B

- 1

991

C -

199

3D

- 1

999

E -

2002

F -

2004

G -

200

7H

- 2

008

G

E

G

18

Judy Estrin

Robert Ackerman

Peter Norvig

Rich Friedrich

Norman Winarsky

Entrepreneur

Allegis Capital

Google

HP

SRI International

It’s very hard for companies to think long-term enough to invest in research. But

they need to anyway.

If you’re a company thinking you can invest your way into some cheap Silicon

Valley research, prepare to have your pockets picked.

There are lots of smart people at Google, but they

don’t do much “blue sky” open-ended research. Except when they do.

It may not be like the old days when Bill and Dave

were still around, but basic research continues at the

company they founded.

Internal R&D is over-rated, and often capable of doing

more harm than good.

“Open Innovation” is an exciting new way to think about research, assuming it’s not just a fancy way of cutting the R&D budget.

You thought it was just a bunch of dorm room hackers, but more and more scientists are hanging out shingles at Facebook.

The path from basic research to a big company is circuitous, which is why it is so seldom-traveled.

The great American research system has produced many successes, and we tamper with it at our peril.

How a traditionally-organized research division keeps up with new trends in research.

Keck Graduate Institute

Facebook

Stanford, VMware

UC Berkeley

Joel West

Cameron Marlow

Mendel Rosenblum

David Patterson

Microsoft

Peter Lee 19

25

22

28

31

Voices from Silicon Valley

34

40

37

43

46

19

As new managing director of Microsoft

Research Redmond, Lee heads up

one of the computer industry’s few

remaining research operations

patterned after the labs of an earlier

era - autonomous and setting its own

lab direction.

Microsoft

A Proud R&D Traditionalist

PeterLee

20

How is it organized?

Research areas are like departments in that they have twenty-

five to fifty researchers and research engineers. And the

research areas are organized around a broad, major direction

in computing. For example, machine learning, visualization

and user experience, and large scale data and analytics.

What do you think of some of the newer ideas that people are talking

about - things like prizes or crowd sourcing or open innovation, etc?

Oh, I think it’s incredibly interesting. The potential for what

can be learned through these kinds of experiments is really,

really strong and, furthermore, it’s a way to really engage a

much broader community and increase the idea flow in ways

that are pretty important. And so today in Microsoft Research,

for example, we’re studying research ideas really closely

and they’re very likely to affect a lot of things that we do.

For example, we’re watching very closely the huge amount

of activity in both the academic and enthusiast communities

around Kinect hacks. That is an example of something that just

kind of spontaneously grew up on its own but is a tremendous

potential source of new, innovative ideas.

Would you say the trend is good or bad?

I don’t see it as either/or. The things that Microsoft Research

does for Microsoft, couldn’t be done any other way. We’re a

teeny, tiny part of Microsoft, but our impact on every single

Microsoft product is really significant, and the visibility that

we have within the company is really amazing.

Each department or each division is expected to show impact;

to show scholarly impact and leadership impact in the academic

community. Show impact on our product groups, and show impact in

terms of...really disruptive new technology.

“ “

orange: People often compare Microsoft Research to the Bell

Labs of old. What do you think of that comparison?

You say you demand returns. How do you measure them?

peter lee: There are some valid comparisons. We are an

independent organization, so the business groups don’t tell

us what to do. Culturally, our researchers are motivated differ-

ently, maybe in ways that weren’t as true at Bell Labs. They’re

highly motivated to get their research ideas onto every desk-

top and into every data center in the world. Even in the short

time I’ve been here, I’ve been astounded at the ambition level

that researchers have when they think they have a good and

useful research result.

I talk a lot of about three lanes of basic research. One

lane being a kind of mission-focused research where we’re

reacting to known problems. The second lane being blue

sky research, often in concert with the academic community.

And then the third lane being the uncomfortable search for

disruptions. We try to have equal rewards for all three lanes

of research here, but also demand that we get good impact

in all three.

Each department or each division is expected to show

impact; to show scholarly impact and leadership impact in the

academic community. Show impact on our product groups,

and show impact in terms of progress towards developing

really disruptive new technology.

The take-aways

21

So the huge amount of the innovation in Bing is a direct result

of our partnership with the Bing group. A huge number of the

underlying algorithms that power Microsoft Office. All the

machine translation products. All the cloud services - Azure,

SQL server, Hotmail - have their origins in Microsoft Research.

And we provide services for predictive analytics of software

defects in a huge range of software development projects. It’s

hard to imagine crowd-sourcing those sorts of things.

Having said that, crowd-sourcing and mass globalization

concepts are extremely interesting, and we believe they are

potentially a great source of innovative ideas.

Examples?

There’s a pretty strong correlation between how profitable a company is and how much it spends on research. Do you worry about funding from Microsoft Research being cut in the event company profits decline?

Why do you think basic research has the reputation that you can’t measure its ROI? What is it that the people are not seeing, or are forgetting, when they have that perspective on basic research?

We’re such a small part of the company cost-wise and I think,

at least the attitude right now is that, in fact, our positive impact

on the company relative to our size is pretty enormous. And

so viewed like that it doesn’t seem like there’s much to worry

about. Of course, we want the company to be very successful

and be more successful every day, but objectively I don’t think

there’s an issue about our security here. Except that, and this

is something I also learned at DARPA, people doing basic

research everywhere in this country have certain anxieties

about society’s understanding and tolerance about basic

research. And so our researchers here aren’t immune to that.

There’s probably a cyclical nature to this. Within the company

right now, the perspective on Microsoft Research and the

value of Microsoft Research are possibly at an all-time high.

And so it’s a very good time. But I’m a realist also. I understand

that these things come in cycles.

Even when you’re at a

company with a long-term

commitment to research,

it’s nice to have the occasional

hit on your hands to keep the

top bosses happy.

“ The things that Microsoft Research does for Microsoft, couldn’t be

done any other way... our impact on every single Microsoft product

is really significant, and the visibility that we have within the

company is really amazing.

“

22

There is no issue related to the

management of R&D more popular at

the moment than “Open Innovation.”

Joel West knows as much about it as

anyone, making his commentary on its

occassional misuse worth nothing. A

consultant and business professor at

the Keck Graduate Institute, he is also

co-editor of Open Innovation:

Researching a New Paradigm from

Oxford University Press.

JoelKeck Graduate Institute of Applied Life Sciences

“Open Innovation” and Its Discontents

West

23

orange: As far as big ideas in business, Open Innovation seems

in fashion, a kind of This Year’s Model.

What is new about the idea? Or is it just a useful name for something

people were already doing?

joel west: Academia has its own fashions, and this is a

particularly hot area, in the way that the “Resource-Based

View of the firm” was hot 20 years ago. If you look up “Open

Innovation” on Google Scholar, you see that there are

thousands of papers. It’s come from nothing in 2003 to where

it is today.

Some people call it the “Old wine in new bottles issue.” Some

people say this is a practice that’s been going on, and it’s just

been given a new name. Certainly, what’s true is that much of

this was already going on.

But there are major differences from the past. One is that

in the Open Innovation approach, the firm is agnostic to the

sources of innovation.

To be neutral about whether the technology comes from

inside or outside is a culture shift for any large, multinational

corporation. In the past there’s been an arrogance at many

large industrial corporations, in which they assume they know

better than anybody in the world. Open Innovation forces

firms to consider outside technologies, rather than saying,

“We have to invent it if it’s going to be something great in the

market.”

A related issue is the recognition by a company that

not all the smart people in the world work for them. That

recognition, I think, is new.

Once upon a time, you could have said that the smartest

people in computing worked at IBM.

But then came the fragmentation of the computer

industry with the PC revolution. You saw manufacturing

and product development going offshore and the Internet’s

dissemination of information, and open source software.

And all of a sudden, people realized that the idea that any

company - even the greatest company in the world - could

have a monopoly or a preponderance of knowledge in an

area just isn’t plausible.

A lot of companies are cutting research budgets these days, and one

can imagine Open Innovation appealing to them as a way of getting

research on the cheap. To what extent is that part of the allure of it?

Do you think the phrase is being overused?

Like sponsoring research at universities?

That has certainly contributed to the interest in Open

Innovation. When I go and talk to people out there, a lot of

the interest is from companies who either A) want to get rid of

their R&D people or B) just gutted their R&D department and

want to know what they should do next.

We don’t have a term for it, but there is an Open Innovation

equivalent of “greenwashing.” Greenwashing is where people

wrap themselves in claims of environmental-friendliness, but

don’t change their actual practices to make their products

more marketable.

When I use Google to see how corporations use “Open

Innovation,” I’d say only about a third of it is really legitimate;

the rest of it is just people want a buzzword to make

themselves seem more innovative and more trendy.

In many cases, when they appoint a VP for Open

Innovation, there is an attitude change and they really are

being more collaborative. At other times, it’s just a new name

for something they’ve always done, and they’re just calling it

something else.

They could have a universities relations arm. They could have

ecosystem management or technology sourcing or technology

IP procurement. They could have, on the other side, a patent

licensing office. Normally, the Open Innovation Officer, VP or

Senior VP or whatever, the Director of Open Innovations, is the

person bringing innovations into the firm. They usually don’t

give that title to somebody who’s trying to find markets for

existing technology.

...the firm is agnostic to the sources of innovation. To be neutral about whether

the technology comes from inside or outside is a culture shift for any large,

multinational corporation.

“ “

The take-aways

24

Like many things in

business, Open Innovation is

part-real, part-hype. The real

opportunity from Open Innova-

tion is when a company is open

to all good sources of ideas.

The hype comes with companies

that hide behind the term as a

euphemism for cutting

research budgets.

Do you think Open Innovation is a good idea, in and of itself, or is it

just something that companies in an era of diminished budgets are

forced to resort to out of necessity or expediency?

The fundamental argument of Open Innovation is that your

R&D operation needs to have competition the same way that

any other aspect of your company needs to have competition.

The vertically-integrated company - where everything is done

in-house and we always use our in-house janitor, our in-

house printing press, our in-house HR manager - that is not

the way business is done today. Apple sells things through

its stores, but it also sells things through other stores. Apple

makes some of its own parts, but it also sources things

outside where it doesn’t have the scale or the technology.

Really, what Open Innovation is saying is that firms ought

to be aware of what the best technology is for anything relevant

to their line of business. They shouldn’t automatically assume

that they do it in-house, nor should they automatically assume

that they do it outside. Instead, they do need to monitor the

state of the art of what’s going on outside the company, to be

able to say when necessary, “Look, this part of our technology

is just not state-of-the-art.”

The fundamental argument of Open Innovation is that your R&D

operation needs to have competition the same way that any other

aspect of your company needs to have competition. Apple makes some

of its own parts, but it also sources things outside where it doesn’t

have the scale or the technology.

“

“

25

The only “research problem” most

people associate with Facebook

is figuring out how to add enough

servers to keep up with its user base.

But Facebook is starting to learn the

lesson of many tech companies; that

if you are in it for the long haul, you’ve

got to start planning for it.

CameronFacebook

A Hot Start-up Begins Its R&D Rite of Passage

Marlow

26

How do you distinguish research from engineering? Couldn’t

someone be listening to everything you’re saying and say, “That’s

not research.”

Can you give me an example of work you’re doing at Facebook that

someone in the academy would look at and say, “Yeah, that would

be legitimate computer science research if it were happening in my

department.”

The types of questions we’re answering are as fundamental

as any academic question could ever be.

The social interactions on the Internet, and on Facebook

in particular, are at a level of detail and scale that haven’t

existed before. They allow us to answer questions about

social interaction and social forces that we’ve never been able

to answer, even though some of these questions are as old

as the discipline of sociology. Of course, the fact that these

questions exist doesn’t mean we’re going to answer all of

them; there isn’t perfect alignment in everything we do. But in

the day of a typical engineer, we need to answer questions like,

“What is the average size of a person’s personal network?” or

“How does that affect the way they use the product?”

I could go on for an hour. We have a lot of interesting work

that’s being done with taking our code base and compiling it

into C. The compiler community is very interested in HipHop

(Facebook’s internally-developed compiler). The issues that

we have with data center usage put us among a very small

number of companies facing issues that are central to the

future of computing. We work a lot with academics on these

problems. And not just from computer science, but also the

social sciences.

We all come from like a very

Internet-friendly research

background, and it’s kind of

expected that if we came to

Facebook, we’d be doing some

kind of research.

The social interactions on the

Internet, and on Facebook in

particular, are at a level of detail

and scale that haven’t existed

before. They allow us to answer

questions about social interaction

and social forces that we’ve never

been able to answer...

orange: So why exactly does Facebook need researchers?

Don’t you guys just sit around and keep growing?

How often are people surprised when they hear about Facebook

doing research-research as opposed to engineering?

cameron marlow: I wouldn’t say that we’re “Facebook

Research,” with proper nouns, but Facebook does a lot of

research. We’ve hired a number of people and published a

number of papers. A lot of researchers coming out of graduate

school, especially those interested in corporate research

labs, want to do great work, but also want to have an impact

on people’s experience. The image I’ve tried to present for

research at Facebook is one where we work as closely with

the product as possible, because the problems we face are

some of the most interesting problems that exist.

I guess I may be a little biased, but I don’t think anyone’s

surprised that I’m doing research, or the people that I work

with are doing research.

We all come from like a very Internet-friendly research

background, and it’s kind of expected that if we came

to Facebook, we’d be doing some kind of research. The

publishing model here is a little different than some other

companies. In other research operations, there is an emphasis

on the number of papers researchers publish, and the talks

they give. Whereas here, the papers and talks are the gravy.

Your real work is working on problems. So there is a bit of

a different incentive here. People really want to make users

more happy. I think it attracts a different type of researcher.

The take-aways

27

How would a traditional researcher find working at Facebook?

Do you think people pay more attention to your work

simply because you’re at Facebook?

What do you think of the way research used to be done at big

technology companies?

I think in a traditional research lab, I would build the prototype,

and then I would show that prototype to a product team, and

then over the course of months, I’d report to them on my

progress at developing my idea into a real product.

Here at Facebook, the code base is available to everyone.

You have your sandbox to work in, so instead of working on

a prototype, you’re working on a prototype that’s actually

connected with the product. When you’re done and people

have seen it and they give you the thumbs-up, you commit

your code. You don’t wait around for some product team to

give you a blessing and build it themselves. If you know how

to do it, just do it yourself. People at Facebook are actually

making changes to the core product, which may not be part

of the DNA of other companies.

We have a PR department here that likes to put us in front of

reporters, and even though the types of things we generate

are on the academic side of things, they tend to be of

great interest to the world. So almost every paper that we

publish is turned into a much bigger communication about

Facebook than just the simple record of the fact that we

published a paper.

The model I used to think about the standard corporate

research labs is that the company was like a planet, with the

lab being in orbit spinning around the planet, and in case the

planet implodes, there’s some chance that this other heavenly

body would spin off and allow the company to continue on.

But I don’t think that actually ever happened. I can’t think of

a time when an AT&T Labs or a Xerox Park fundamentally

changed the way that the core company operated.

Successful young compa-

nies have many advantages,

not the least of which is that

researchers can get an idea

into the hands of millions

of users in not much longer

than it takes to test the code.

I can’t think of a time when an

AT&T Labs or a Xerox Park

fundamentally changed the way

that the core company operated.

Here at Facebook, the code base is available to

everyone. You have your sandbox to work in, so

instead of working on a prototype, you’re working on a

prototype that’s actually connected with the product.

“ “

“ “

28

Rosenblum, as an associate professor

at Stanford University, did the original

theoretical work that led to VMware, making

him one of the handful of Silicon Valley’s

rare entrepreneurs who was able to grow

some basic research into an enormously

important company.

MendelVMware, Stanford

On Being A Basic Research Poster Boy

Rosenblum

29

orange: First off, do you even agree that the work you did on

virtualization would qualify as basic research?

In what sense was that unfamiliar terrain back then?

mendel rosenblum: I was involved with a group of

people trying to build a super computer, a very, very, large

machine. But I wasn’t really interested in scientific computing,

so I was trying to find out if we could use it for something

else, like running a whole enterprise’s worth of computation.

That’s how we stumbled on the idea. If you look at the original

papers, we were talking about running a bunch of virtual

machines with modern computing environments on a single

machine. We didn’t know it at the time, but the vision turned

out to be the right one.

The idea of virtual machine monitors was actually invented

by IBM in the late 1950s and early 1960s. But as PCs became

more common, it pretty much died as a research idea. I’m

an experimental system builder, and so I propose systems,

and when I demonstrate them, I try to build prototypes of

them. In the original paper, I re-launched virtualization. It was

a mixture of old and new ideas. Some of the newer stuff, like

the transparent memory sharing, hadn’t been done before,

Did you have anything commercial in mind at first?

We were going to do a virtual machine monitor for one of

the big servers like Digital Equipment or HP. That’s what our

research was on: building software for the big servers. And

so we went and talked to the companies making them, and

the first one said, “Why don’t you just come and join us and

do it as an employee?” But that didn’t sound very interesting

to me.

Why not?

So why the PC?

Who pushed you to do the company?

Oh, I don’t know. I guess we had the idea that you’re not

going to get like rich and famous building up a big company

if you do it as a team inside a big corporation. I remember

talking to a vice president at Digital Equipment Corporation,

and he named these examples of projects that they had

nurtured inside DEC and then spun out. As far as I could tell,

they were all disasters.

The nice thing about trying to do it for the PC was it was pretty

clear we didn’t depend on anybody. The PC was opening up,

and we knew what the hardware did, so it made it more of a

tractable problem to do it as an outside company.

I had two graduate students, and they’re actually the two

graduate students who have helped found VMware, Edouard

Bugnion and Scott Devine. They’d been sitting around

watching the Yahoo guys, David Filo and Jerry Yang, take

off and become famous. So they immediately said, “Can we

commercialize it?” It wasn’t an ideal time for me, because I

was coming up for tenure. But I talked to (Stanford University

president) John Hennessy about it, and he told me he had

started MIPS when he was coming up for tenure, so he didn’t

see a problem.

I was trying to find out if we could use it for something else, like

running a whole enterprise’s worth of computation. That’s how we

stumbled on the idea.

“ “

The take-aways

30

Graduate students are

motivated by many things,

including reading about how

rich and famous other graduate

students have become.

Academics who hope to also

do well in the marketplace

would be wise to surround

themselves with them.

It’s definitely true that when you’re

in a position at a company, and you’re

looking at where to spend your

money, everything is focused on the

short-term of the company. With

research, the hardest thing for people

is that you don’t really know how

you’re going to benefit from it.

Some people might say you weren’t living up to your own ideals.

How research-friendly was VMware in the days when you had

something to say about it?

It’s definitely true that when you’re in a position at a company,

and you’re looking at where to spend your money, everything

is focused on the short-term of the company. With research,

the hardest thing for people is that you don’t really know how

you’re going to benefit from it. So if you can’t really figure out

what it’s going to do, chances of it being funded aren’t going

to be very promising.

Well, VMware viewed itself as an innovative company, and

one of the ways we hoped to stay ahead of the competition

was to out-innovate them. So there was a focus on trying to

keep innovation going. But basic research - research where

it’s not obvious how it’s going to apply or benefit - that just

wasn’t done.

The problem is companies are focused on how something is going to return

to the bottom line. Basic research is getting pretty rare. “ ““

“Okay. Fast forward to the world today. Do you think research is

sufficiently appreciated at tech companies in Silicon Valley?

The problem is companies are focused on how something is

going to return to the bottom line. Basic research is getting

pretty rare. People have to get funding from product groups.

That means you have to convince a product group that what

you’re doing will help them at some point in time.

31

Dave Patterson is an embodiment

of the elite of traditional academic

computer research. He is a professor

at UC Berkeley, known for his

microprocessor architecture work

with Stanford’s John Hennesey, and

recipient of numerous prizes and

recent president of the ACM.

DavidUC Berkeley

The Best Days of Traditional Research Still Lie Ahead

Patterson

32

How much of the research now being done at Berkeley can only be

done in academia, and how much of it could be done someplace in

the industry?

So what are the advantages that we in academia have? We

get brilliant people from all over the world who don’t know that

things can’t be done. The U.S. university system is the best

in the world. If you ranked the top ten universities, probably

eight of them would be here.

So why in the world would you want to leave out a really

bunch of brilliant people? Sure, industry does a lot, but

industry often, especially today, has a shorter term focus.

We can take this longer term. And we also have an extremely

valuable by-product, in that we produce the next generation

of leaders.

Of course, what’s happening is that start-ups play a more

important role in the field than when I got here. But start-ups

aren’t supposed to be doing research. If a venture capitalist

thinks a startup is proposing to do research they back off.

It’s up to programmers to deliver on Moore’s Law now. We can put more

transistors into chips, but we can’t turn it into performance unless we solve

one of the hardest problems computer science has ever faced.

orange: There are people who think that computers and IT

have gotten so mature that we can now leave it to private industry

to fund R&D the way we once did with telegraphs or radios or TV.

What do you think?

What are some examples of research that don’t just cure headaches

but open up new possibilities??

david patterson: That’s just crazy. We’ve just scratched the surface of information technology. In my career, I can look back to the things I learned when I was a student that my own students laugh at when I tell them.

It’s going to be the same way when they’re older. One

example involves all the security problems we have with the

technology we invented. If we have solved those problems,

why is there WikiLeaks? The weakness of our technology is a

major security threat to this country. It’s embarrassing that it’s

so vulnerable. People are relying on programs like Windows

NT for safety and control of critical systems. Those of us who

have been in this field for a while are embarrassed by it.

Technology is amazingly cheap and amazingly fast. But

there are still real big holes. You might call this the “headache

model” of funding research.

Well, kind of a nerdy thing is the Parallel Computing

Challenge. Easy-to-program parallel computing is the hardest

problem computer science has faced. We’ve been working

on it continuously for 50 years. We’ve been trying to work on

making progress on it. Everything that Intel ships has parallel

cores in it. The whole industry has bet its future that we’re

finally going to solve it.

So we are forced to transform the whole information

technology stack to make parallelism a first class citizen.

It’s up to programmers to deliver on Moore’s Law now. We

can put more transistors into chips, but we can’t turn it into

performance unless we solve one of the hardest problems

computer science has ever faced.

It’s not a controversial statement. Lots of start-up

companies have tried and failed with the bet that they could

finally make easy-to-write parallel programs. There’s a Dead

Parallel Computer Society filled with names of companies

funded by venture capitalists.

Technology is amazingly cheap

and amazingly fast. But there

are still real big holes. You might

call this the “headache model”

of funding research.

“ “

The take-aways

33

What are some other favorite unsolved problems, where we still need

deep, basic innovation?

When did the start-ups begin playing more of a role?

It seems that there could be a lot of downside to that.

There are also good reasons to work with industry,

aren’t there?

We’re actually making advances through statistics and

machine learning on what might be called “augmented

intelligence” instead of “artificial intelligence.” Like cars that

couldn’t crash. If we were to set a national program to make it

difficult for cars to crash, the savings in lives, in medical bills,

in dollars would be just phenomenal. There have been enough

advances in a lot of the fields that it is not a ridiculous goal to

have a national program to dramatically reduce traffic deaths

and traffic accidents. We have lots of pieces of the technology

to make this real, not science fiction.

Well I should say today graduate students are more tempted

today with some new and exciting technologies rather than

finishing their PhDs, so they decide to go and maybe start a

company.

...today graduate students are more

tempted today with some new and

exciting technologies rather than

finishing their PhDs, so they decide to

go and maybe start a company.

If this was happening all the time it’d be hard for the university.

If 75% of project stopped, it would be hard to be able to

complete research projects at universities. But it’s not that

common. It happens, but it’s manageable.

Yes. It’s easy in academia to pick problems that most people

don’t care about. And that’s an advantage of interacting with

industry. Pasteur invented pasteurization so he could drink

milk. He made a fundamental contribution to science, but

also solved a problem. When we interact with industry, we

understand the problems more like eight or ten years down

the road. We start working on them so that five or six years

later, we have something that can be impactful.

Patterson is a strong

advocate for the old-fashioned

idea of basic research as an end-

less frontier, where the answers

to today’s questions give rise to a

new set of queries that lay the

groundwork for the next

generation of inquiry.

We get brilliant people from all

over the world who don’t know

that things can’t be done. The U.S.

university system is the best in the

world. If you ranked the top ten

universities, probably eight of them

would be here... Of course, what’s

happening is that start-ups play a

more important role in the field than

when I got here. But start-ups aren’t

supposed to be doing research. If a

venture capitalist thinks we’re doing

research, they back off.

“

“

34

Estrin has been an entrepreneur,

(Bridge Communications) the CTO

of a big tech company (Cisco), a

board member for several major

corporations (Disney, FedEx, Sun

and Rockwell ) and has written

about research (Closing The

Innovation Gap).

JudyEntrepreneur

The Dangers of Taking Research For Granted

Estrin

35

orange: You’ve been on many boards of directors. What have

you learned from that experience about corporate research?

What can companies do?

judy estrin: The average term of the CEO is three to four

years. But basic research happens on a five- to ten- to 15-year

time frame. And so if you’re incentivizing your management

on a two- to five-year time frame, you’re not going to get basic

research. At best you’ll get applied research, or advanced

technology. So one after one - and there are few exceptions

- companies have pretty much taken what they called their

“research” and made it more applied.

What they have said in their defense is, “We’re connecting

our research closer to the customer, so that we will get more

benefit from it.” And it really is a benefit not having labs

isolated, so researchers are more connected and can produce

things that the company can profit from. The disadvantage is

if you are focused on connecting to today’s customers, you

won’t get a very disruptive technology. The customer doesn’t

know what they might need ten years from now.

I don’t think they can afford to do what the corporations of the

past did, meaning build big labs. I’m not even sure that works

anymore. But I think that they can afford to do a few things.

Intel tried for a while what they called Lablets. These were

groups of researchers that were co-located with academic

researchers. They had this very tight relationship with a couple

of universities and then it actually was a very interesting

model for leveraging their resources with academic resources

to further research.

So what is wrong about research moving to the academy?

As a result, the responsibility of research has more and more

fallen to academia. Now, unfortunately, what has happened in

academia is that while there is some great research going on,

but it is not getting enough funding. The scarcity of funding

causes researchers to be more risk-adverse, because they

want to submit grants that are more likely to get funded. So

we miss the wild and crazy ideas, the ones that you want

to have happen, even if they’re going to fail, because in the

research environment you learn from failure. Look at the

opportunities that are out there on the intersection of IT and

nanotechnology and biotechnology, or look at our ability to

start understanding neuroscience better because we now

have the tools. The stuff isn’t getting funded.

But if you’re looking at things like

systems - clean tech, biotech,

complicated hardware - it’s

tougher. There is a gap in between

academic discovery and what a

venture capitalist will fund. It’s

what I call the “lab gap.”

...unfortunately, what has happened in academia is that while there

is some great research going on, but it is not getting enough funding.

The scarcity of funding causes researchers to be more risk-adverse...

“ “ The major problem with research only being in academia

is that you can only go so far in a university lab. The labs

are relatively small. We’ve gotten a little spoiled because

in the software industry you can go directly from academia

to a venture financial company - Google, Yahoo - because

they’re based on an algorithm. But if you’re looking at things

like systems - clean tech, biotech, complicated hardware - it’s

tougher. There is a gap in between academic discovery and

what a venture capitalist will fund. It’s what I call the “lab gap.”

The take-aways

36

Estrin is one of a breed of

business leaders who have

become alarmed at the cur-

rent lack of regard for basic

research, and are using their

positions of influence to call

attention to the issue.

There are people who say, well, Groupon isn’t doing research,

Facebook isn’t doing research, Zynga isn’t doing research. Doesn’t

the success of these companies show that researchers are

decreasingly necessary?

Research has nothing to do with the success of today.

Research has to do with the success over the long term.

So you have no idea where Facebook is going to be five or

ten years from now. I would point to Google as an example

that has always had a little bit more of a commitment to, not

necessarily basic research, but advanced technology. They’ve

always spent some of their money on stuff that isn’t about

today. I think it’s the reason why Google has continued to be

able to grow.

I will point out that Facebook and Groupon and all of

these companies are based on research that was done 20 or

30 years ago. If we hadn’t built the Internet, if Tim Berners-Lee

hadn’t built the Web, if he hadn’t, you know, done some of the

algorithmic stuff, Facebook wouldn’t exist.

Any ecosystem of innovation has three pieces to it. It has

research, development and applications, and you can innovate

in any one of those communities. And they all feed off of each

other. But they have to be in balance. If you cut off research

ultimately innovation in development and application will die.

It may take 20 years, but it will ultimately die.

I will point out that Facebook and

Groupon and all of these companies

are based on research that was done

20 or 30 years ago. If we hadn’t built

the Internet, if Tim Berners-Lee hadn’t

built the Web, if he hadn’t, you know,

done some of the algorithmic stuff,

Facebook wouldn’t exist.

Let me ask you a question about Cisco. They are pretty well known

for their sort of acquisition model of research - meaning they tend to

buy things rather than develop them internally. Do you think that’s a

good strategy?

Cisco used acquisitions for advanced technology and for

furthering innovation. There’s no research in that equation.

When I was there, Cisco did no research. There was no focus

at all on the basic research or the longer term stuff. They did

some funding of people in academia, and did some donations

to networking groups at different universities. But it was

generally not basic research. And I think they have changed

some, because after the bubble burst, they realized that

maybe there weren’t always going to be companies out there

to acquire, and that were some areas where they needed to do

more of their own research. TelePresence is a good example

of stuff that they did more internally, although it started with an

acquisition. One of the disconnects when I was there, I have

this fundamental belief that a leader in an industry needs to

Research has nothing to do with the success of today. Research has to do with

the success over the long term...I have this fundamental belief that a leader in an

industry needs to have some component of this longer term thinking.

“ “

have some component of this longer term thinking. And again

at least when I was there, which was the peak of the bubble,

everything was about time to market, taking advantage

of the market.

37

Ackerman is a venture capitalist, but

more specifically, one associated with

the International Business Forums and

its conference on Corporate Venturing

and Innovation Strategies, for companies

interested in research-oriented alliances

with Silicon Valley VCs.

RobertAllegis Capital

What VCs Won’t Tell You About Silicon Valley R&D

Ackerman

38

That seems pretty cynical.

The history of corporate venturing, with few exceptions,

has born that out. They get in, and they get out. Every two

years, you’ll have new people in place. You’ll have changes

in strategic direction. Corporate priorities will ebb and flow.

When the markets get competitive, the top corporate guys

look at the venture program - which is usually generating

losses - they say ‘Who got us into this? Fire him. Get us out

of it.” They forget everything they learned. But five years later,

they’ll decide to start all over again.

But a lot of companies do investing that’s strategic rather than im-

mediately financial. Aren’t the rules different for them?

If you’re not making money, then you’re not safe. “Strategic”

means “losing money.” You go into a corporation, and

everything that’s “strategic” is losing money. When budgets

come under pressure, you do not want to be “strategic.” You

want to at least be carrying your own weight.

That’s the problem for these programs. Many times in the

venture ecosystem, the time that it takes to really begin to

realize those strategic benefits is outside of this cycle. And so

it’s very, very difficult to build a sustainable program

The history of corporate venturing, with few exceptions, has born that

out. They get in, and they get out. Every two years, you’ll have new

people in place.

“Strategic” means “losing

money.” You go into a

corporation, and everything that’s

“strategic” is losing money.

When budgets come under

pressure, you do not want to

be“strategic.”

orange: What do companies need to know if they want to

invest in Silicon Valley?

So what should he say instead?

robert ackerman: Venture capital is a club, a tight little

club, organized around managing risk. As a VC, I want to invest

with people I’ve invested with in the past, because I know

how they’re going to be there when times are good and when

times are tough. We syndicate. We share information. I’ve got

everybody calibrated.

But a corporation is different. The corporation walks in

and says, ‘Hi, I’m from a big global company. Perhaps you’ve

heard of us. We operate in 110 companies around the world.

We have 140,000 employees. We have a market cap of $42

billion. And we’re inviting ourselves to your party.”

The polite venture capital response to them is, “Fantastic,

let’s find things we can collaborate on.” But what the venture

capitalist is actually thinking is ‘Okay, what can I sell to this

guy? How do I pull money out of his pockets and use it for

whatever I need to get it into?”

The more truthful kind of response would be, “Who cares that

you’re a big global company, because you may be here today,

but you’re going to be gone tomorrow. You’re reassuring me

of your commitment, but you’re corporate direction is going

to change. You’re not a long term player. And so you’re

going to be a tool of convenience for me in my ecosystem.”

“ “

The take-aways

39

So why should companies bother with Silicon Valley in the first

place?

It sounds like it comes in waves.

When you’re seeing the future for the first time, there’s a leap

of faith required. As venture capitalists in Silicon Valley, we’re

in the business of inventing the future. Yet that’s very difficult

to do, and that’s why venture capitalists sometimes have a

herd mentality, like, “If one of those is good and successful,

we need 400 of them.” But it’s the guy who does it the first

time, who sees it for the first time, who has the conviction to

pursue that vision and organize people around him - how rare

that type of person really is.

It’s fascinating. We’re seeing a wave of this again. It’s picking

up again, and they’re all going to go make the same damn

mistakes. They’re just going to do it again.

Venture capital is a game where you’ve got to be inside

baseball. The corporate guys can’t integrate into our world

because they’re slow, they’re pondering. There’s a whole

culture around these large corporations. That’s why they don’t

innovate quickly. We want to be able to reach out to them, on

our terms, where we can leverage them. But they can’t get in

the way of what we’re trying to do.

Companies who think they

can use occasional venture

investments to gain access to

Silicon Valley research will be

greeted with open arms. Then,

the trouble will begin. They

should keep an eye on their

wallets.

The corporate guys can’t integrate into our world because they’re slow, they’re

pondering. There’s a whole culture around these large corporations. That’s why

they don’t innovate quickly.

When you’re seeing the future for

the first time, there’s a leap of faith

required. As venture capitalists in

Silicon Valley, we’re in the business

of inventing the future. Yet that’s

very difficult to do,

“ “

40

As director of research at Google,

Norvig has his stars aligned: he

is at a highly profitable company

with a propensity for hiring Ph.D.’s.

Google doesn’t have the same sort

of separate research operation that

Microsoft does, as Norvig explains.

PeterGoogle

Staying in Touch With Google’s Academic Roots

Norvig

41

orange: Google has the reputation for being crammed with

Ph.D.’s, but not having a separate, Bell Labs-style research outfit.

What is Google’s philosophy about research?

Does anyone do purely speculative, “blue-sky” research at Google?

peter norvig: I think we have a bit of an identity crisis, and

we’re still thinking of ourselves as a startup. At startup, you

get a bunch of Ph.D.’s together, along with the non-Ph.D.’s,

and everybody just pitches in. If you have to invent something,

you invent it. But you don’t really make distinctions, saying,

“This guy’s doing (Research), but that guy’s doing (Product

Development).

That’s the way it’s always been at Google.

We have a research effort, with a couple hundred people out

of our 20,000. But 90 percent of our Ph.D.’s are not on the

“Research Team.” They’re on the “Engineering Team.” Some of

their time is spent inventing things that haven’t been invented

before. If you want to call that “Research,” fine. If you want to

call it “Product Development,” well, that’s fine too.

I would say mostly no. But let me qualify that. We don’t have

anybody whose job it is to write papers, or to prove theorems.

Everything you do at Google should be useful for something.

But we may not know right away how that useful thing will fit

into a product.

This goes back to the start of the company. People always

say Google’s core competency was “Search.” But when it was

a research project back at Stanford, Larry and Sergey they

didn’t think they were doing “Search.” Their idea was, “Hey,

there’s a lot of really cool stuff on the Internet, and we should

get ourselves a copy of it and see what we can do.” And so

they scrounged around for enough computers, and then they

started experimenting. And it was only later that they said,

“You know, the thing that we can actually do is ‘Search.’”

Some people tell me that university research isn’t as interesting as it

used to be, that it’s gotten closer to what companies do.

Some companies idea of research seems to be that they can crowd-

source a question on the Internet whenever they need to know

something. What do you think of that approach?

I think it’s a sign of maturity. In the 60s or 70s, nothing had

been done yet, so it was easy to do something brand new. But

now, to build something exciting takes longer. You can’t just

go out and say, “Well, here’s a brand new field nobody’s done

before, and here are some results.”

I think there’s still interesting work going on in universities.

Although I also think there are some things you can’t do in

universities, and that’s one big reason why I’m not in the

university now. You need such a huge number of computers

just to put together a lab. As an assistant professor, I might be

able to afford five grad students. That would put me at a big

disadvantage, compared to the industrial teams.

When new things are made, they’re built on the back of a

long tradition. So in order to build that something new, you’ve

got to have the expertise to do it. Some of that expertise you

can get by just reading, or by talking to people that you’re

not paying. But you really have to understand it at a deep

level. And I don’t think that comes for free. You can’t just say,

“Oh, I want to play in this new area I know nothing about it, so

I’ll put a question on some question site and expect to get a

good answer.”

...We have a research effort, with a couple hundred people out of our

20,000. But 90 percent of our Ph.D.’s are not on the “Research Team.”

They’re on the “Engineering Team.”

...Everything you do at Google should be useful for something. But we may

not know right away how that useful thing will fit into a product. “

“ “

“

The take-aways

42

The difference between

“research” and “engineering”

at Google is often just a matter

of what is written on your

business card.

A lot of people look around and see companies like Groupon and

Twitter and FourSquare, where they seem to be creating enormous

value without any kind of traditional research.

There are a lot of good ideas out there, and you can put

together things in lots of ways. Google has concentrated on

things that require a deeper level of expertise. But there are

other applications where the algorithmic part is trivial, and the

value added is being able to connect people to each other in

just the right way. We couldn’t do that without the technology

and the networks that we have now. But if you can find places

where you can capitalize on that infrastructure, great.

I read a comment to the effect that the Google self-driving car shows

why government shouldn’t be involved in research, because Google

was able to make a car all by itself, without needing any help from

the government.

What is Google’s relationship with academia like?

That team that works on the Google car sits in the same

building as me, and when I look out over their cubicles, I see

the three guys from Carnegie Mellon who were sponsored by

government work; a guy from MIT, who has been sponsored

by the government; and a couple of guys from Stanford, also

government funded. All of them have government dollars

piled up underneath them.

We really feel like we have to play nice with academics. We

owe them a debt, and we want to continue to have interactions

with them. We want to hire their students. We want to bring

in their interns over the summer. We want to have visiting

professors here. We want to fund their work and we want

to participate in conferences. We have to go to conferences

and talk to them and read their papers. But we also have to

publish papers, because if we were sitting in the back of the

room saying nothing, they’d stop trusting us. So you’ve got to

participate as a full citizen in that community.

...when I look out over their cubicles, I

see the three guys from Carnegie Mel-

lon who were sponsored by govern-

ment work; a guy from MIT, who has

been sponsored by the government;

and a couple of guys from Stanford,

also government funded. All of them

have government dollars piled up un-

derneath them.

...we have to play nice with academics.

We owe them a debt, and we want to

continue to have interactions with them.

We want to hire their students.

I also think there are some things

you can’t do in universities, and

that’s one big reason why I’m not in

the university now... As an assistant

professor, I might be able to afford

five grad students. That would put

me at a big disadvantage, compared

to the industrial teams.

43

Friedrich is the director of the

Strategy and Innovation Office

at Hewlett-Packard, a company

often described as an example of a

venerable tech leader being forced

by a changing world to cut back on its

legendary commitment to research.

RichHewlett-Packard

HP Is Still About More Than Just PCs and Printers

Friedrich

44

So how many people at HP are doing basic research?

Give me some example of things from research that you were able to

add that Dell wasn’t able to.

HP’s Vinay Deolalikar got a lot of attention recently for solution he

offer for the P=NP problem. What do you say to a shareholder who

asks, “What is this P=NP thing, and why are we paying this guy all

this money to solve something that’s not going to help the stock?”

Has HP been able to resist short term financial pressures as

successfully as you’d like? Are there times when you think it should

have pushed back harder?

We have about 500 researchers, so if you take a third of that,

that would be around 150 or 170. It’s not always easy to

compute that number, because a person may spend two thirds

of his or her time doing something that’s really fundamental,

but a third working on applying it. So you’re never 100%

dedicated to one particular category.

So the couple things that really caught on fire were the Instant-

On DVD and Instant-On CD players for the laptops. It used to

be that you had to boot up Windows before you could run a

CD player or watch a movie. But we built instant technology

by paring down Linux, hosting a mini DVD player and a CD

player. Another example is high end audio, which we’ve had

in the last couple years and which has made a big impact.

Lighter weight laptops, higher resolution screens - there’s

been a series of things.

I’d paraphrase the response that Maxwell gave to Queen

Victoria when she asked about Maxwell’s equations, and how

these could possibly be relevant to England. His response

was something to the effect, “Your majesty, how do we judge

the potential of a baby?” To be a little more concrete, I would

say that if we were doing something related to astrophysics,

then I think a shareholder would be right to wonder why we’re

doing it. But when you’re dealing with algorithms, you’re never

quite sure how they’re going to influence what you do later on.

That’s a question for the EVPs and the CEO. But I would say

from my position as a director inside the organization I think

HP has done a reasonable job of resisting lure of the sirens of

the quarterly Wall Street report.

...our Memristor, which has the

potential to revolutionize all

of memory and all of storage,

came about because of

investments that Dave Packard

asked the company to make

back in the

mid 1990s.

orange: So why do you always hear “Too bad about HP,

because they’re doing a lot less of the kind of research that made

them famous.”

So give me some examples of basic research, the kinds of things

that people are doing at HP.

So the amount devoted to basic research has doubled, but it’s a

bigger piece of a smaller pie?

rich friedrich: So three and a half years ago when

our Senior Vice President for Research, Dr. Prith Banerjee

joined, I would say that the total research investment of HP

was making only about 10 or 15 percent focused on basic,

fundamental research. He’s upped that considerably, so that

today we’re about one third basic research.

If you look at where we’re making a lot of fundamental

investments today, one of the big areas is nanotechnology,

especially as applied to next generation computing

infrastructure. In fact our Memristor, which has the

potential to revolutionize all of memory and all of storage,

came about because of investments that Dave Packard

asked the company to make back in the mid 1990s.

We also have a lot of photonics work. We’re looking

at how ions behave at the nano level. And we’re making

significant investments in fundamental research around

information analytics, which is all very deep mathematics and

science applied to how you extract out of data something

that’s meaningful.

Right.

The take-aways

45

You’ve described how government’s sponsorship of research was

becoming more short term oriented. What is the implication of that?

After the dot com phenomena, a lot of professors left

universities, many in hopes of becoming billionaires. As

the country kind of recovered from that event, there was

a lot more pressure applied by groups like the National

Science Foundation and the National Institutes for Health

for universities to find a corporate sponsor for some of

their research. It’s typically meant that timeframes had to

be pulled in.

Fundamentally, the United States has become very short

term focused, even on our fundamental research, which

means no one’s working on that computing architecture that’s

going to be really important in the 2020 to 2030 timeframe.

You effectively drain the current profit pool with no way to

invest to get your company to the next one.

HP’s total research budget may

have shrunk, but the company

doesn’t want anyone thinking its

commitment to basic

research is undiminished.

I would say that if we were doing something related to astrophysics, then

I think a shareholder would be right to wonder why we’re doing it. But

when you’re dealing with algorithms, you’re never quite sure how they’re

going to influence what you do later on.

Fundamentally, the United States

has become very short term focused,

even on our fundamental research,

which means no one’s working on

that computing architecture that’s

going to be really important in the

2020 to 2030 timeframe.

Let me give you a really concrete example. About a

decade ago one of our senior fellows, Bernardo Huberman,

did a project which allows executives in a company to do

a better job of predicting things such as revenues. They’ve

shown that they can take what a decade ago were considered

very abstract algorithms and turn them into something that’s

very powerful for an enterprise. They’re using them now to

predict things, based on analyzing the chatter that goes on

Web 2.0 sites.

“ “

46

46

Winarsky is a vice president at SRI,

the research outfit often best-known

for its pioneering work with the

personal computer. Much of SRI’s

work today involves performing

research for other companies.

NormanSRI

Is Your R&D Operation Really Your Friend?

Winarsky

47

orange: SRI does research for other people. But what’s wrong

with a company doing research itself, the way Bell Labs did?

That implies that, to the extent companies might be spending less on

research, that’s not necessarily a bad thing.

That sounds frightening.

norman winarsky: You’re asking, “Why can’t a company

have an R&D institution that feeds its research into the

company for products and services, but can simultaneously

be open to outside opportunities for software, hardware, and

the like?” The reason is that I’ve never seen an example of

that. It’s natural for an R&D organization to consider outside

R&D to be competitive with their own approach. It takes new

and innovative management to change the structure, so that

open innovation can be possible.

Research equates in my mind to invention. So if you agree

that invention can come from anywhere, and that you’re open

to that, then I think spending money on invention is good, to

the extent that you have a goal. But one needs to have a goal

of innovation, not research: You have to start with a market

problem in mind. Companies generally have a problem with

what we call at SRI the Innovation Valley of Death.

Funding for basic research definitely seems to be in decline, but

do you think it’s been empirically proven to be not profitable? Do

you know for a fact that Bell Labs wasn’t, for example, specifically

profitable for AT&T?

I haven’t read a study that states that, and probably because

no one would like to see the results, nor would they share

them. But I’ve talked to executives of major companies,

hundreds by now, and in a private room after having had a

scotch, they’d admit that the cost of R&D is greater than the

profits of it.

...It’s natural for an R&D

organization to consider outside

R&D to be competitive with their

own approach. It takes new

and innovative management to

change the structure, so that open

innovation can be possible.

Research institutions generally

like doing research that is

three to five years away from

product or service. That’s what

a researcher loves to do.

...But I’ve talked to executives of major companies, hundreds by

now, and in a private room after having had a scotch, they’d admit

that the cost of R&D is greater than the profits of it.

“ “Research institutions generally like doing research that is

three to five years away from product or service. That’s what

a researcher loves to do. But companies want delivery into

the marketplace within the time frame that’s going to affect

their stock price and their revenue. That means 12-18 months.

So researchers want to stay three or more years out, but

companies want to stay no more than 18 months out. Between

those is the Valley of Death. Companies need to make sure

that they’re either investing for the crossing of that valley, or

that someone else is. This becomes an issue of the board of

directors of that company. You know, they have to oversee

the long-term success of the company.

The take-aways

48

You won’t hear any lamenting

about the good old days of Bell

Labs from SRI. If you run the

R&D lab inside a company, you’ll

probably consider them

competition. Which in fact is

their whole point.

So is there any research left to do?

What about the patent officer that said that you should close

the patent office? This has never been a better time for

innovation in IT. At SRI, we have three major themes. One

theme which started with Siri (later bought by Apple) is for the

virtual personal assistant. We absolutely believe that being

able to personalize and contextualize a person’s interaction

is deep technology. Another example of what SRI’s working

on is augmented reality. Twenty years ago, SRI created the

company that did what you see on the walls of football games,

all those ads and so forth and on the field. Augmented reality is

now entering a new generation of helping other experiences,

such as games and shopping. The third big theme is cyber-

security and privacy.

Those are just three examples. There hasn’t ever been a

better time for research that will deliver into the marketplace,

which is what we call innovation, than right now.

Many people are struck by the number of new tech companies today

that seem to be enormously valuable, but which seem to not be

based on any sort of research.

What we’re discovering is that a company’s value proposition,

or its sustainable competitive differentiation, which we at SRI

call the “Golden Nugget,” does not have to be technology -

by any stretch of the imagination. What Zynga, Facebook,

Groupon and the others have done is spectacular. They’ve

understood there is a new type of business proposition,

which is equally valuable, or even more valuable, than

technology. The business proposition is an untapped market

opportunity that comes from having an audience connected

with technology. We’ve invented some power tools, namely

the Internet and broadband connections and iPads and

mobile phones, and at that point, technology is no longer

the limiting problem.

What Zynga, Facebook, Groupon and the others have done is spectacular.

They’ve understood there is a new type of business proposition, which is

equally valuable, or even more valuable, than technology.

There hasn’t ever been a better

time for research that will deliver

into the marketplace, which is what

we call innovation, than right now.

“ “

Orange SIlicon Valley Project Team

Natalie Quizon

Mark PlakiasGeorges Nahon

Pascale Diaine Minesh Govenji

References and images in this publicationPage 4, Microscope Stage 3, by Larry Darling, available under Creative Commons at http://www.flickr.com/photos/tncountryfan/5543913413 Page 9, Gas Mask, Broad Arrow, Zero, available under Creative Commons at www.flickr.com/photos/noodlefish/3871148611/in/set-72157622187321634/Page 11, 1 Foray, Dominique, Hall, Bronwyn H. and Mairesse, Jacques, Pitfalls in Estimating the Returns to Corporate R&D Using Accounting Data (September 1, 2007). CDM Working Papers Series CEMI-WORKING PAPER-2007-003. Available at SSRN: http://ssrn.com/abstract=1427754Page 11, Illustration: Experimental modes of transistors, 1953. From Bell Laboratories, Arthur Gregor, 1972.Page 18 and 19:Peter Lee, http://www.flickr.com/photos/msr_redmond/5515738115/Page 18 and 22: Joel West, http://www.kgi.edu/Images/Faculty/west_joel.jpgPage 18 and 25: Cameron Marlow, http://www.flickr.com/photos/cameronfactor/3624252776/in/photosof-cameronfactor/Page 18 and 28: Mendel Rosenblum, http://www.ece.utoronto.ca/aboutus/dls1/Rosenblum.htmPage 18 and 31: David Patterson. http://www.eecs.berkeley.edu/department/EECSbrochure/c6-s5.htmlPage 18 and 34: Judy Estrin, http://www.ischool.berkeley.edu/newsandevents/events/dls20090415Page 18 and 37: Robert Ackerman, http://www.allegiscapital.com/team-ackerman.htmlPage 18 and 40: Peter Norvig, http://pn.smugmug.com/gallery/1677561#27898845_VZvrjPage 18 and 43: Rich Friedrich, http://senseable.mit.edu/futurecities/speakers.htmlPage 18 and 46: Norman Winarsky, http://www.triplehelixconference.org/keynote-speakers/norman-d-winarsky.html

Lee Gomes’ acknowledgements:

I’d first like to thank Georges Nahon of the Orange Institute, whose support and ideas were

central to this project. In addition to the people profiled in this report, I’d also like to thank the

following, who made significant contributions of one kind or another. It should not be assumed

that anyone who helped with the project is in agreement with its conclusions.

Mark Boroush, National Science Foundation Jules Duga, BattelleHossein Eslambolchi, 2020 Venture PartnersJason Freidenfelds, GoogleVirginia Gold, ACMRichard R. John, Columbia UniversityBronwyn Hall, UC BerkeleyRobin Hanson, George Mason UniversityLillian Hoddeson, University of IllinoisDavid A. Hounshell, Carnegie Mellon UniversityBarry Jaruzelski, Booz & Co.Ed Lazowska, University of WashingtonLeslie Lamport, MicrosoftBaruch Lev, NYUAndrew Odlyzko, University of MinnesotaRichard H. Van Atta, Georgetown UniversityChrissy Vaughn, Waggener Edstrom

Lee Gomes, who lives in San Francisco, has written about research, technology and Silicon

Valley companies for two decades, much of that time at the Wall Street Journal.

Publication design by Minesh Govenji

�

her code Engendering Change in the Silicon Valley

Orange Labs San Francisco

Other Publications from Orange Silicon Valley

JA

VA

C

++

P

HP

V

ISU

AL B

ASIC

C#

P

YTHON P

ERL OBJECTIVE-C HTML5 JAVASCRIPT ACTIONSCRIPT GO

LISP C R

UB

Y

Orange Labs SF

the coming of

Porous Enterprise

1

social c:ommerce

$%

From New Age to New Edge (with Orange Institute) Sensor Networks (with Orange Institute)

Her Code: Engendering Change in Silicon Valley

DevNation: State of the Developers Nation

The Coming of the Porous Enterprise

Social Commerce: The Future of Shopping

1

sensor networks

Orange Institute

as the New Growth Opportunity

thoughts

planet

communities

bodies

Copyright 2011 – Orange Silicon Valley