The Star-Ledger Archive
COPYRIGHT The Star-Ledger 2004

Date: 2004/03/21 Sunday Page: 001 Section: NEWS Edition: FINAL Size: 2040 words

Talent leak drains AT&T think tank

Once a bastion of cutting-edge research, it's lost its star power


When AT&T Labs was carved from Bell Labs in the 1995 breakup of AT&T, the telecom giant set lofty goals for its new research arm.

"Our mission, in my view, is to invent the future of communications," proclaimed Alexander "Sandy" Fraser, who pushed to create AT&T Labs.

Today, many of AT&T's top scientists still chase that dream - somewhere else. They strive to invent the future in the shiniest ivory towers and hottest tech companies, from MIT to Microsoft, from the Pentagon to Google.

Some 200 scientists - nearly half the core research staff - were let go from AT&T Labs in Florham Park in January 2002 amid sweeping corporate cuts throughout AT&T. Since then an all-star collection of researchers has bolted from the labs.

The fate of AT&T Labs mirrors changing fortunes at AT&T, an American icon squeezed by bad investments and bad timing. More importantly, some scientists say, it raises tough questions about the direction of industrial research and America's future as an innovator.

At AT&T Labs, the brain drain is so severe, observed Michael Kearns, now at the University of Pennsylvania, that his former employer's motto should be "404 Not Found" - the error message that greets many searches on the labs' Web site.

Defectors point to the loss of esteemed colleagues, cuts in long-range research and restrictions on travel, media contacts and publication of scholarly articles. The place has had three vice presidents of research within the past year.

For some researchers, the last straw was having to pay their own way to present scientific papers at prestigious conferences. For others, it was the elimination of free espresso and bottled water at the leafy Florham Park campus, once the estate of Vanderbilt descendants.

Yet many remember the brief heyday of AT&T Labs, during the euphoria of the Internet boom, as the most thrilling time of their careers. For them, the exodus is a tragedy.

"We had a national gem," said Avi Rubin, who exposed flaws in electronic voting systems last year as a faculty member at Johns Hopkins University.

"To see it melt away is very painful," said Andrew Odlyzko, who sensed trouble brewing in 2001 and left to head a digital technology center at the University of Minnesota.

While turmoil at AT&T Labs is a bonanza for places like Columbia University and the federal Defense Advanced Research Projects Agency, scientists say it underscores the decline of "blue-sky" research - science for science's sake - at Bell Labs in Murray Hill, IBM, General Electric and the Xerox Palo Alto Research Center.

Gone from AT&T Labs, or nearly so, are groups highly regarded for their long-term studies in artificial intelligence and machine learning, network security and cryptography, algorithms and theoretical computer science, and statistics. AT&T research operations in Cambridge, England, and at the University of California, Berkeley, are gone, too.

The National Science Foundation says federal support for basic science has waned, as well, since 1980.

"It's an open question where the next big ideas and discoveries will come from," said Paul Saffo of the Institute for the Future. A former adviser to AT&T Labs, Saffo warned that corporate America's "relentless race for short-term value is killing our future . . . AT&T Labs was a national crown jewel - and it's been terribly devalued."

"If you're focusing on research that's short-term, to impact products in a year or two, there are all kinds of world-changing discoveries that you simply miss," said Maria Klawe, president of the Association for Computing Machinery and dean of engineering at Princeton University.

Princeton has cherry-picked at least two AT&T Labs scientists since 2002; Klawe interviewed another this month. The university even created an institute for materials sciences last year specifically to "help fill a national void" left by declining resources of industrial research labs.

For its part, AT&T says fierce competition has forced a shift from basic science to business-driven research. Projects now must improve the bottom line within months, not years, as AT&T morphs from a phone company to a supplier of business networking services. When AT&T finally shed about $100 billion of cable TV and wireless ventures - disastrous investments meant to satisfy Wall Street during the tech boom - it also shed prime areas for research.

"We are playing to win," AT&T Labs President Hossein Eslambolchi told industry analysts in February.

Labs spokesman Michael Dickman called the downsized AT&T Labs a "lean, mean networking machine," focused on ensuring the reliability of AT&T's vast data network. Partnerships with universities will play a bigger role going forward, he said, declining to tout anyone still at AT&T Labs.

"We had the names, the celebrities. That was then. This is now. We don't have people like that. Even if we did, it goes against our strategy to highlight them," said Dickman.

Among prominent names to bail recently:

Lorrie Faith Cranor. Named one of the world's top 100 young innovators last year by MIT's Technology Review, the Internet privacy expert left in December to teach at Carnegie Mellon University.

Matt Blaze. The cryptographer has exposed flaws in everything from common locks to the Clinton administration's "Clipper Chip"; he left in December for the University of Pennsylvania.

Peter Shor. A pioneer in "quantum computing" - he showed how it might crack the most secure encryption someday - the MacArthur Fellow quit last summer to become a math professor at MIT.

Others have gone to Israel's Weizmann Institute of Science and to Microsoft Research - maybe the only shop rich enough to support basic research as Bell Labs did in the Cold War, when AT&T's telephone monopoly paid the bills.

Not surprisingly, AT&T Labs patterned itself after Bell Labs, birthplace of the transistor and winner of 11 Nobel Prizes.

More startling was AT&T's decision to launch a new lab at all.

When AT&T was split up in 1995, Lucent Technologies inherited its equipment-making business - and Bell Labs. Sandy Fraser, a genteel British researcher from Bell Labs, helped convince AT&T, now strictly a services company, it still needed a lab.

"It was very courageous in a way for a services company to embrace the idea of having its own research organization," said Ron Brachman, an expert in artificial intelligence who followed Fraser from Bell Labs to AT&T Labs. (Both left in 2002 Fraser started a research firm, and Brachman joined the Pentagons Defense Advanced Projects Research Agency.)

Lucent kept most of the chemists, physicists and materials scientists. AT&T vied for Bell Labs computer scientists, mathematicians and, especially, speech recognition experts. It was messy. Lucent kept the libraries of Bell Labs; AT&T got the librarians.

"We like to think we stayed with it, the parent lab," said Ron Graham, a Bell Labs math star whose jump to AT&T Labs inspired others to follow. He left for the University of California, San Diego, in 1999.

AT&T Labs named its Florham Park headquarters for Bell Labs legend Claude Shannon. Researchers even tried juggling on a unicycle, as Shannon once did, said Michael Littman, now a Rutgers professor.

About 80 percent of research involved projects with an eight- to 10-year window. Now, turnaround targets are 18 to 24 months, said Dickman, the AT&T Labs spokesman.

Early on, the diverse talent mix surpassed the most elite schools, former researchers said. Brainstorming was encouraged in lounges called "bump spaces." Mathematician Eric Rains (now at the University of California, Davis) joined the quantum computing group thanks to a discussion overhead during his job interview. Cryptographer Rebecca Wright, now at the Stevens Institute of Technology, said rounding up dozens of computer security experts used to be easy at AT&T Labs.

They were galvanized by Fraser, who spoke of a digital "Renaissance Network," and David Nagel, a former Apple Computer executive who once designed cockpits for NASA. Nagel, the first president of AT&T Labs, now is CEO of PalmSource Inc.

"I used to think I had the best possible job," said Avi Rubin, who envisioned a long career at AT&T Labs. He compared it to a great university - without the hassles of grading exams and chasing grants.

Research was eclectic. AT&T Labs tried teaching computers to learn from mistakes. Researchers designed intelligent scheduling devices, smart antennas and wireless delivery of local phone service.

They dabbled with Internet delivery of music before Napster, and with Internet video phones. Computer scientists were widely quoted in debates about instant messaging, privacy and security. That last topic struck close to home David Smith unleashed the Melissa virus in 1998, when he was a contractor for AT&T Labs in Florham Park.

About 300 researchers work there now, roughly half the number from the late 1990s, estimated Dickman. Another 6,000 people - mostly in development, not research - work at a mammoth complex in Middletown. Thats down by about 1,500 from its peak, Dickman said. A hundred or so researchers are based in Menlo Park, Calif., and about 30 more staff a lab in Nice, France, he said.

As the air whooshed from the tech bubble, AT&T slashed spending on research and development from $550 million in 1999 to $254 million in 2002, according to Schonfeld & Associates, a business research firm.

Free espresso was among the casualties.

"Thats real penny-pinching," said Jim Reeds, snapped up by the Institute for Defense Analyses in Princeton after the 2002 layoffs. Lorrie Faith Cranor, who paid her own way to deliver a paper at a Florida conference last spring, said life grew confusing at AT&T Labs.

"There was a lot of pressure to tie everything to the immediate im pact on the business. At the same time, they told us they understood the importance of doing research not directly tied to the business," said Cranor, whose husband Chuck, a networking researcher, also left AT&T Labs.

Rubin said "baby sitters" from public relations were assigned to all media interviews. He said he struggled for permission to publish a research paper about the Postal Services vulnerability to cyber at tacks. When a co-worker was laid off, he resolved to split.

Peter Shor said he felt so isolated after all the departures that he had to leave. "Nowadays, I dont know what the mission is" at AT&T Labs, he said.

Others said AT&T Labs had no choice but to downsize.

"We soldiered on as well as we could, quite competently. And we got mugged - by Wall Street," said Bjarne Stroustrup, inventor of a popular programming language. Stroustrup left AT&T Labs last year for Texas A&M but retains ties to the labs. He doubts universities can pick up the slack from corporate research labs.

"They dont have the size or the culture or the reward mechanisms or the management experience. Universities dont operate on the scales that Bell Labs and AT&T Labs did" in focused areas, said Stroustrup.

The president of AT&T Labs insists the organization is helping AT&Ts bottom line. Technology from the labs foiled the Slammer worm last year, Hossein Eslambol chi told industry analysts in February. He praised advances in speech recognition, natural language understanding and artificial intelligence for automating customer service. He promised more advances in high-speed data over wireless networks and power lines, and technology to aggregate voice and e-mail messages.

Eslambolchi, who holds four job titles, compared AT&T Labs to a big league ball team.

"It is the talent of the players ... that differentiates teams," he told the analysts. "AT&T has the winning players, and we are playing to win. ... This is our story, and we are sticking with it."

NOTES: "We soldiered on as well as we could, quite competently. And we got mugged - by Wall BJARNE INVENTOR OF A POPULAR PROGRAMMING LANGUAGE



-----Original Message-----
From: Michael Kearns []
Sent: 10/31/2005 4:59 PM
Subject: RE: reporter request

PS Pls as usual send me a pointer to your piece when finished? Also, do you have a pointer to an online version of
your earlier article on AT&T Labs that does not require a login to view? Thanks.
Meant to mention to you in the context of your current article and just the topic in general: were you aware that
Yahoo has a large and ambitious new plan for a (relatively basic) research program spanning multiple labs? The
head of a new lab opening soon in NYC is none other than Ron Brachman, recently departed from DARPA and
formerly back at AT&T/Bell Labs. The estimated figure for the overall research budget is 100M annually, which
would put it about the size of AT&T Labs in the good days of the late 90s. Thought you might be interested if
you did not already know.

-----Original Message-----
Sent: Monday, October 31, 2005 12:17 PM
Subject: RE: reporter request

It's going to be one of those weeks!  I've got a phoner from 12:30 to 1. Sorry for the e-tag...
-----Original Message-----
From: Michael Kearns []
Sent: 10/31/2005 11:59 AM
Subject: RE: reporter request

I've got a brief window at 12:30 today (half an hour from now); if you can do that ring me on 215.898.7888?
-----Original Message-----
Sent: Monday, October 31, 2005 11:54 AM
Subject: RE: reporter request

Hi Michael--
Looks like I may be away from the office on assignment Wednesday. Any chance you can spare a few minutes anytime today or Tuesday afternoon?
-----Original Message-----
From: Michael Kearns []
Sent: 10/30/2005 9:30 AM
Subject: RE: reporter request

Kevin --- sure, be happy to talk about your next piece. What about Weds sometime?

-----Original Message-----
Sent: Friday, October 28, 2005 6:20 PM
Subject: RE: reporter request

Hi again,  Michael-
I will send you my AI story via separate email.
Now I am writing a piece about the future of Bell Labs. Can we talk about that sometime next week?
-----Original Message-----
From: Michael Kearns []
Sent: 10/21/2005 9:58 AM
Subject: RE: reporter request

Hey Kevin, good to hear from you. What's the DARPA project? Is it the Grand Challenge?
Coming to a total dollar figure for AI funding over all history would be a pretty tricky enterprise ---
others might be able to do a better guesstimate than me --- but I would conservatively estimate
in the hundreds of millions and quite likely billions of dollars. Part of the difficulty is that AI is
a pretty broadly defined and changing subfield of CS (unlike, say, theory), so even defining the
scope would be hard. But I'm pretty comfortable with the figures just given, crude as they are.
More locally there is more one can say, I think. During the heydays of AI and machine learning
research at AT&T Labs and Bell Labs, I would estimate spending in the single-digit millions
annually. I myself ran an AI/ML research from 97 to 01 that had roughly 15 headcount, so the
total loaded cost of personnel, equipment, travel, etc. was probably 2M+ annually right there, and
there were a couple of other groups I would also count as AI research. NEC probably had comparable
size groups during the same time period, Siemens a smaller effort.
NJ was indeed once a hotbed of AI research. In all honesty, during the 90s the group I ran
(which was started by Ron Brachman and run by Henry Kautz before me) was widely regarded
as the single strongest AI/ML group in the world, bar none, or at least right there with the best.
This includes not only industry but all of academia. The main point, which I think you are probably
familiar with, is that AT&T/Bell Labs, once they decided to focus on an area, were willing to build
much more deeply than academic departments will. If an academic department hires a faculty
member in AI one year, hiring another the following year has greatly reduced odds.
And I would indeed say that the AI prominence in NJ has been largely eradicated by the demise of
AT&T Labs, Bell Labs, NEC Research, etc. There are some strong people in academia in NJ (many
of them formerly from the various labs as you know) but in terms of sheer critical mass and energy
it's nothing compared to the 90s.
AI as a field more broadly, however, is thriving in my view. In retropsect the concentration of
excellence in NJ was perhaps an anomaly, and now the talent is largely in academia and
geographically dispersed.
In terms of funding, as you may know all of CS research has experienced a squeeze of late that
many are concerned about. However, given this state of affairs it certainly seems to me that
AI is getting at least its fair share of funding, and hopefully the squeeze is a cyclical phenomenon.
Let me know if this helps.

-----Original Message-----
Sent: Thursday, October 20, 2005 11:12 AM
Subject: reporter request

Hi Mike--

I am writing a story about an A.I. project at DARPA.  I am trying to get a handle on two things: How much money has been spent on A.I. research over the decades, and how much of the research went on in NJ.  (Bell Labs, AT&T Labs, Siemens, NEC, etc.)

Any idea on the $ figure?

Was NJ once an A.I. hotbed?  Does much A.I. research continue there--or is the "A.I. winter" still in force?



Kevin Coughlin
Technology Writer
The Star-Ledger
Star-Ledger Plaza
Newark, NJ 07101
voice:  973-392-1763
fax:      973-392-5845

CONFIDENTIALITY NOTICE: This e-mail may contain information that is privileged, confidential or otherwise protected from disclosure.  If you are not the intended recipient of this e-mail, please notify the sender immediately by return e-mail, purge it and do not disseminate or copy it.

CONFIDENTIALITY NOTICE: This e-mail may contain information that is privileged, confidential or otherwise protected from disclosure.  If you are not the intended recipient of this e-mail, please notify the sender immediately by return e-mail, purge it and do not disseminate or copy it.

CONFIDENTIALITY NOTICE: This e-mail may contain information that is privileged, confidential or otherwise protected from disclosure.  If you are not the intended recipient of this e-mail, please notify the sender immediately by return e-mail, purge it and do not disseminate or copy it.

CONFIDENTIALITY NOTICE: This e-mail may contain information that is privileged, confidential or otherwise protected from disclosure.  If you are not the intended recipient of this e-mail, please notify the sender immediately by return e-mail, purge it and do not disseminate or copy it.

CONFIDENTIALITY NOTICE: This e-mail may contain information that is privileged, confidential or otherwise protected from disclosure.  If you are not the intended recipient of this e-mail, please notify the sender immediately by return e-mail, purge it and do not disseminate or copy it.