Calif. ó IBMís California research lab sits atop a
green hill here, 15 miles south of downtown San Jose.
arenít any signs that suggest if you drive up the
narrow road that wraps around the hill youíll find a
research facility at the top. No signs that the research
center is home to a Fortune 500 company. No signs ó
even inside ó that the company once dominated the
personal computer industry.
decades in the spotlight as a hardware-centric firm
selling PCs, servers and mainframes, the 105-year-old
tech giant has made a dramatic shift into a realm that
few understand: cognitive computing. Deep within the
apps we use, the food we eat, the medicine we take and
the medical diagnoses we make, youíll find traces of
its Almaden research lab, there are no obvious signs itís
there. Look carefully, though, and the work of IBM ó
particularly from its Almaden lab ó is everywhere.
cognitive computing business ó which includes
artificial intelligence, machine learning, algorithms
and analytics ó accounted for 35 percent of the
companyís $81 billion in revenue last year. Itís the
fastest-growing segment at IBM, where overall sales are
declining. Jeffrey Welser, director of IBMís Almaden
Research Center, said itís "the main focus for
IBM in terms of growth."
work at Almaden played no small part in this.
hundred researchers work at the facility developing
artificial intelligence, algorithms and the chip sets to
support faster, more powerful self-learning systems. It
feels less like other Silicon Valley tech campuses,
where hoverboards and ping-pong tables are common
sights, and more like a college campus without the
students. The hallways are quiet and clean. The tenured
professors ó in this case, the IBM scientists and
engineers ó code away in their discreet offices.
is still a major player in servers and mainframes ó
its technology continues to power banking and airline
reservation systems. It also has people working in areas
such as cybersecurity and the cloud. But to understand
why it sees cognitive computing as its future, it helps
to go back to 2011, when IBMís artificial
intelligence, Watson, beat human players on the game
"Jeopardy" isnít like winning at chess. An
earlier IBM artificial intelligence, Deep Blue, beat the
world chess champion in 1995 by computing every possible
move and picking the best option. But for Watson to even
play "Jeopardy", it had to learn natural
language, understand riddles and answer questions in
coherent sentences. Stuff humans do.
had been researching artificial intelligence since the
1970s. But the Watson victory was a major breakthrough:
If IBM could teach Watson everything it needed to know
to win a complex game show, what else could it be
of it is showing that a computer can do things beyond
crunching numbers," Welser said. "The human
mind cannot crunch numbers very well, but it does other
things well, like playing games, strategy, understanding
riddles and natural language, and recognizing faces. So
we looked at how we could get computers to do
after the "Jeopardy" matches, IBM researchers,
including those from its Almaden lab, taught Watson to
read patent databases and medical journal abstracts. In
drug research, any given molecule can have as many as
100 synonymsóbrand names, generic names, different
chemical strings. They taught Watson how to identify
those. They also taught it context.
it reads the word Ďsleepyí in a document, it has to
understand whether sleepiness is the cause, effect, side
effect, or if itís what the drug is trying to stop or
achieve," Welser said.
feature, which is part of the Watson Discovery Advisor,
is now used by researchers in the pharmaceutical
BEGIN OPTIONAL TRIM)
not all that Watsonís been taught.
cognitive power that goes into Watson has also given IBM
a seat at the food safety and genomic research table.
company partnered with Mars Inc. last year to create a
consortium that, in a bid to improve food safety,
studies the fingerprints of bacteria, fungi and viruses
and how they grow in different environments. Much of
that research is being pioneered at Almaden.
has partnered with the Memorial Sloan Kettering Cancer
Center to enable medical researchers to use Watson to
get detailed diagnostics and treatment options based on
the latest cancer research.
radiology, Watson can go through thousands of images to
help doctors find MRI scans that are relevant to a
can even be found in recruitment and marketing tools,
and dating apps. The artificial intelligence can trawl
through social media accounts, analyze the language
people use and determine their personality traits.
test its efficacy ó and just for fun ó researchers
at Almaden have fed Watson scripts from the "Star
Wars" movies to see if it can figure out the
personality traits of characters. On the neuroticism
scale, C-3PO ranks high, while Jedis tend to rank low.
On the anxiety scale, C-3PO also comes in on top. When
it comes to assertiveness, Yoda outranks everyone.
about understanding people better at a deeper
level," said Rama Akkiraju, an engineer at Almaden
who works with the Watson group. "So if Iím
trying to sell you something, I could know ahead of time
what kinds of things youíre interested in, and Iíd
only offer you the things relevant to you."
bad for a computer. Except Watson isnít technically a
computer, itís software. It helps to think of Watson
as a collection of algorithms stored in the cloud that
can do different things. They can be mixed and matched
to solve a problem ó like an army, in which every
soldier has a different skill.
in these armies of algorithms, powered by faster and
more efficient chips, that IBM believes its future lies.
move hasnít been completely painless, though. IBM has
spent the last decade going through job cuts and revenue
dips. Today it employs 377,000 people around the world,
down from nearly 450,000 in 2011. In April it reported a
sales drop for the 16th consecutive quarter to $18.7
an interview with Bloomberg, IBMís chief financial
officer, Martin Schroeter, said its cognitive solutions
business will have "a fair bit of ramp in order to
make them sizable and start to punch above their weight
in terms of overall growth rate."
also faces increasing competition in the cognitive
computing space. Google, Apple and Facebook have all
invested in artificial intelligence, with algorithms
assisting users with search, image and face recognition,
and voice recognition (Hello, Siri!).
smaller start-ups making photo and music apps are
teaching their algorithms to recognize images and songs,
and learn from peopleís usage habits.
just like IBM saved companies from having to build their
own computers and servers, itís now hoping it can turn
a profit saving companies, scientists and researchers
from building their own cognitive solutions. Need help
with data crunching? Genomic sequencing? Personality
matching? Thereís a Watson for that.
itís quite content to stay in the background this
time, being the invisible backbone for consumer-facing
companies. Like its Almaden lab, its contributions are
hard to spot. Look carefully, though, and youíll find