Artificial Intelligence: Learning

Through Interactions and Big Data

An IBM Redbooks® Artificial intelligence (AI) has been formed as a field of computer
Point-of-View publication by IBM science for understanding and building intelligent entities since
Academy of Technology the 1950s. As with most technologies, the concept has
undergone many transformations over time. In particular, the
Authored by Koichi Takeda, IBM Research field of AI has experienced several cycles of expectations,
Tokyo, and Tamiya Onodera, IBM Research commitments, and then disappointments, sometimes referred to
Tokyo
as AI winters, where interest in the technology has waned only
to then peak again at a later time. Some might argue that these
cycles of conflict have caused inconsistencies that have made
forward progress difficult in the AI field. However, recent
practices of data-centric approaches (for example, machine
Highlights learning and probabilistic reasoning) are beginning to
demonstrate the true capability of AI to solve practical
AI approaches combined with analytics decision-making problems.
techniques can provide practical
approaches to the challenges of creating The concept of AI is often thought of as having two dimensions:
smarter solutions.
򐂰 One dimension emulates the human thought process or
򐂰 The current mainstream business simulates intelligent behavior (such as cognitive computing
analytics research and solution and brain-inspired systems).
development along with the
򐂰 The other dimension achieves human or ideal performance
convergence of AI with machine
(such as question-answering and natural language
learning techniques will continue to
understanding capability as demonstrated by IBM Watson
underpin high-value, human decision
and its Deep Question-Answering technology).
support solutions.
򐂰 Cognitive systems can combine natural This dual nature of AI challenges organizations with how to best
language processing, hypothesis approach AI modeling solutions. Conventional AI approaches
generation and evaluation, and dynamic often tackle knowledge acquisition and reasoning in terms of a
learning for a powerful, fast, and set of ad hoc rules and manually crafted weights for computing
intelligent problem solving. confidence. As a result, acquiring and maintaining consistency
of rules and improving the accuracy of the reasoning can be a
򐂰 To effectively apply intelligent problem
challenge when the size of the knowledge base grows to solve
solving using AI solutions, you can
real-world problems. This knowledge bottleneck for expert
extend the capabilities of IBM Watson
systems is often perceived as an inhibitor for building scalable
and its DeepQA technology architecture
industrial solutions.
to other domains.
However, AI approaches remain central to intelligent problem
solving. Unlike precisely defined mathematical problems, where
algorithms for computing complete or approximate answers can
be applied, many real-world problems are often underspecified,
ambiguous, or incompletely defined and understood. AI
techniques turn out to be effective for modeling the problem and
building practical solutions.

One important aspect of modern AI is that it now embraces

machine learning techniques for building statistical models and
making statistical reasoning and prediction. Recent advances in
statistical approaches along with the availability of a massive
amount of data (big data) have been remarkable. These
approaches are driving enhancements for high-value business
analytics.

Redbooks
© Copyright IBM Corp. 2013. 1
According to Bernard Meyerson, Chief Innovation doctor can better determine possibilities for treatment.
Officer, IBM, the DeepQA technology of IBM Watson “is In turn, better data might help reduce unnecessary
just a first step into a new era of computing that’s going treatments or tests, thus improving patient outcomes
to produce machines that are as distinct from today’s and reducing overall costs, for the patient, the practice,
computers as those computers are from the and ultimately, the healthcare industry, while also
mechanical tabulating devices that preceded them.”1 making processes more efficient.
As organizations embrace AI, understanding the
methods to approaching intelligent problem solving is In another example, a cognitive computing assistant,
central to gaining an advantage with AI solutions. such as IBM Watson Engagement Advisor, can help a
business to better serve its consumers. This
technology allows brands to crunch big data and
transform the way that they engage clients in key
AI and business analytics drive functions, such as customer service, marketing, and
smarter solutions sales. IBM Watson Engagement Advisor can learn,
adapt, and understand a company’s data, enabling
Machine learning and statistical or predictive analysis, customer-facing personnel to assist consumers with
in many cases, boils down to business analytics, which deeper insights more quickly than previously possible.
is a core component of contemporary AI. Thus,
Other examples of adapting AI technologies to other
business analytics are central to the AI-related
industries might be the deployment of speech
activities that build a smarter solution. When you
recognition and voice command interpreters to
approach methods to use AI technologies, consider the
smartphones, consumer electronics, and automotives
following AI-related activities:
and the application of computer vision to robots.
򐂰 Adapting AI technologies to other industries
򐂰 Applying cognitive computing to mind-like A system that generates hypotheses can
computational models
recognize different probabilities of various
򐂰 Converging AI and business analytics
outcomes. It can “learn” from successes and
򐂰 Incorporating operational intelligence into expert
failures and can improve future responses.
integrated systems

Adapting AI technologies to other

industries Applying cognitive computing to
mind-like computational models
An AI program can use natural language capabilities,
hypothesis generation, and evidence-based learning to Cognitive computers move beyond simple processing
produce a better outcome. Researchers are now of data and begin to learn through experiences. By
looking at advances in technologies to leverage AI in applying cognitive computing, models can find
other industries. Facts strongly suggest that the correlations, can create hypotheses, and then learn
question and answer architecture behind the AI from the outcomes. This process is fundamental
technology that made IBM Watson possible can be research for building the coherent and unified model of
applied to other domains. If quality information sources mind-like computation. Potentially, the resulting model
are available and a reasonable set of sample questions can be applied to solve massive sensory information
and answers for the domain exists and if the deep processing and cognitive problems.
linguistic analysis for the source language is available,
this type of technology can adapt well to other Researchers at IBM have been working on a cognitive
industries. computing project called Systems of Neuromorphic
Adaptive Plastic Scalable Electronics (SyNAPSE). By
For example, the core technology behind AI can help reproducing the structure and architecture of the
transform the way a doctor interacts with a patient. By brain—the way its elements receive sensory input,
feeding data to an AI program, such as IBM Watson, a connect to each other, adapt these connections, and
transmit motor output—the SyNAPSE project models
1
http://asmarterplanet.com/blog/2012/12/the-ibm-5-in-5-our- computing systems that emulate the brain’s computing
2012-forecast-of-inventions-that-will-change-the-world-wit efficiency, size, and power usage without being
hin-five-years.html programmed.

2 Artificial Intelligence: Learning Through Interactions and Big Data

One of the earlier achievements in the SyNAPSE For more information about BAMS, see:
project is the (cat-scale) cortical simulations (with 109 http://www-07.ibm.com/in/research/bams.html
neurons and 1,013 synapses) that were run by the Blue
Gene/P supercomputer with 147,456 CPUs and Convergence of AI and business analytics approaches
144 TB of total main memory at the Lawrence can further extend the scope of analytics into
Livermore National Laboratory. structured and unstructured information, human
behavioral patterns, and interactive learning of
Cognitive systems can transform how knowledge. The convergence of AI and analytics
promises to model intelligent decision processes.
organizations think, act, and operate.
Learning through interactions, they deliver
Incorporating operational intelligence
evidence-based responses that can drive better
into expert integrated systems
outcomes.
Today’s applications are complex. They involve a
diverse set of hardware and software components that
interact with each other horizontally and vertically.
Converging AI and business analytics Optimizing, deploying, configuring, and managing
these components require an enormous amount of
Recent empirical AI approaches deploy analytical manual effort by human experts. This tuning process is
methods—called machine learning—by assuming a extremely laborious and error-prone. Automating this
statistical model for capturing uncertainty in the given process in a cost-effective manner belongs in the realm
problem. Machine learning works by clearly defining of applications for artificial intelligence.
business goals, exploring data, and validating and
deploying models. It relies on domain and data In the mid 2000s, IBM researchers successfully applied
expertise, rather than analytical skills. This method control theory to software systems, resulting in two key
estimates a set of parameters of a model (the features in a commercial product:
probability) from an ample amount of observed data 򐂰 Utility Throttling of DB2 v8.2
that approximates the problem space. 򐂰 Adaptive Self-Tuning Memory of DB2 v9
Machine learning techniques are often theoretically These features made it easier to configure and
sound, and various mathematical properties are known maintain databases for all market segments.
and used to formulate the given problem. Machine
learning approaches are inspired by the advances and In mid 2012, IBM delivered PureSystems, a breed of
success of Business Analytics and Mathematical expert integrated systems that capture and automate
Science (BAMS) research. BAMS conducts active what experts do in order to make every part of the IT
research in several areas of analytical and lifecycle easier. PureSystems uses a semantically rich
mathematical sciences such as optimization, language, developed by IBM Research, to define its
algorithms, stochastic analysis, statistics, data mining, building blocks. The knowledge that formerly sprawled
predictive modeling, and simulation. across hundreds of pages of manuals is now distilled
into a formal representation.
BAMS current projects include a diverse range of
industry domains, such as retail and banking service As the technology integrates further, expert integrated
lines, including Supply Chain Management (SCM) and systems will evolve by incorporating more usage data
Human Capital Management (HCM) and solution for improving performance and consumability, which
areas, such as Smarter Cities and Smarter Energy. will eventually be known as operational intelligence.
Examples of projects that BAMS has executed include
capacity utilization, hiring and training, strategic
planning and budgeting, workforce scheduling, A grocer who is stocking store shelves can use
contingency planning, dynamic supply chain an instrumented glove designed for a cognitive
optimization, customer targeting, smarter water supply
computing system that monitors sights, smells,
management for growth markets, propensity modeling
for cross-sell and up-sell, and so forth. texture, and temperature to flag bad or
contaminated produce.

3
Extending the capabilities of From the healthcare industry, the technology behind
IBM Watson is expected to branch into other industries
IBM Watson and DeepQA that rely on analytic solutions to manage unstructured
data, as illustrated in Figure 1.
To effectively apply intelligent problem solving using AI
solutions, the successful organization must extend the
Wa tson for Watson for Watson For Watson for
capabilities of today’s technologies. IBM Watson He althc are C lient Fina nc ia l Indus try
Engagem ent S vcs.
represents only the beginning of a new way of thinking

Solutions
Advisor Advisor Advisor
about ways that technology can help us live and work Solutions Solutions Solutions

better. IBM Watson and its Deep Question-Answering

In stitutional
C all Center

Knowledge

Retirement
Utilization

Institution
Help Desk
Onco logy

Technical
Diabetes
(DeepQA) technology architecture can be applied to

Banking
Card iac
other domains or in other languages for QA.

For example, healthcare question-answering for

medical diagnosis support is one of the most promising ASK Services DISCOVER Services DECISION Services
areas for IBM Watson. In October of 2012, IBM

Capabilities
announced a collaboration with Cleveland Clinic
clinicians, faculty, and medical students to enhance the
capabilities of Watson’s DeepQA technology for the NLP & Big Data Ana lytic s Cloud Mobile Workload
Machine
area of medicine.2 Watson’s ability to analyze the Learning
Optimized
Systems

meaning and context of human language and to quickly

Plat form
process information to provide answers can help
healthcare decision makers use the huge volumes of
information generated in their industry. Cleveland Clinic Source Model Train Learn
along with IBM recognized the opportunity for Watson
to interact with medical students to help explore a wide Figure 1 IBM Watson products and infrastructure
variety of learning challenges that the medical industry
faces today.

In February 2013, IBM, WellPoint, Inc., and Memorial

Next steps: Leveraging IBM
Sloan-Kettering Cancer Center announced the first Watson and IBM expert
commercially developed Watson-based cognitive
computing breakthroughs.3 For more than a year, IBM
integrated systems
has partnered separately with WellPoint and Memorial
Sloan-Kettering to train Watson in the areas of IBM’s capability in a diverse range of technical fields
oncology and utilization management. During this time, such as QA, cognitive computing, and business
clinicians and technology experts have “taught” Watson analytics with workload optimized systems positions
how to process, analyze, and interpret the meaning of IBM as a unique trusted partner to achieve truly
complex clinical information using natural language intelligent problem solving with optimized platforms.
processing, all with the goal of helping to improve
Your organization can now take advantage of the
healthcare quality and efficiency.
following IBM solutions, which are powered by Watson:
These innovations represent a breakthrough in how 򐂰 IBM Watson Clinical Insights Advisor
medical professionals can apply advances in analytics
Designed for therapy designers, this solution assists
and natural language processing to “big data,”
with efficient trials and reduces time to market with
combined with the clinical knowledge base to create
new cancer therapies. It can help accelerate
evidence-based decision support systems. These
research and insights.
Watson-based systems are designed to assist doctors,
researchers, medical centers, and insurance carriers, 򐂰 IBM Watson Diagnosis and Treatment Advisor
and ultimately to enhance the quality and speed of Designed for oncologists, this solution assists in
care. identifying individualized treatment options for
patients diagnosed with cancer. It can help improve
diagnosis and treatments.
2
http://www-03.ibm.com/press/us/en/pressrelease/39243.wss
3 http://www-03.ibm.com/press/us/en/pressrelease/40335.wss

4 Artificial Intelligence: Learning Through Interactions and Big Data

򐂰 IBM Watson Care Review and Authorization Advisor 򐂰 IBM Watson At Your Service: New Watson
Breakthrough Transforms How Brands Engage
Designed for nurses, this solution streamlines
Today’s Connected Consumers
manual review processes between a physician and
health plans. It can improve decisions and http://www-03.ibm.com/press/us/en/pressreleas
outcomes. e/41122.wss
򐂰 And Now, From I.B.M., Chef Watson, New York
IBM also offers IBM PureApplication System, a cloud
Times, 27 Feb 2013
application platform that is designed to simplify and
accelerate the deployment of new cloud-based http://www.nytimes.com/2013/02/28/technology/
applications. PureApplication System supports a broad ibm-exploring-new-feats-for-watson.html?ref=t
range of needs, ranging from traditional enterprise echnology
applications to mobile, social, and analytics solutions. 򐂰 IBM Watson Progress and 2013 Roadmap
As data grows more and more complex, a simpler http://www.slideshare.net/manojsaxena2/ibm-wa
approach to managing IT is needed. Expert integrated tson-progress-and-roadmap-saxena
systems fundamentally change the IT lifecycle by 򐂰 IBM Journal of Research and Development, “This is
reducing costs, saving time and resources, and Watson”, Vol.56, No.3/4, May-June 2012
speeding innovation for your enterprise. IBM
PureSystem combines the flexibility of a http://ieeexplore.ieee.org/xpl/tocresult.jsp?
general-purpose system, the elasticity of cloud, and isnumber=6177717
the simplicity of an appliance to provide an expert
integrated system that has the ability to transform the
IT lifecycle.

These IBM smarter solutions show how technology can

interact with human experts for learning and enhancing
the accuracy, scalability, and speed of decision-making.
AI approaches, in addition to analytics techniques, can
usher us toward such intelligent systems.

Resources for more information

For more information about the concepts highlighted in
the paper, see the following resources:
򐂰 The Era of Cognitive Systems: An Inside Look at
IBM Watson and How it Works, REDP-4955
http://www.redbooks.ibm.com/redpapers/pdfs/re
dp4955.pdf
򐂰 IBM Smarter Planet solutions
http://www.ibm.com/smarterplanet/us/en/overvi
ew/ideas/
򐂰 IBM Analytics for Healthcare
http://www-01.ibm.com/software/analytics/heal
thcare/index.html
򐂰 The IBM Next 5 in 5: Our 2012 Forecast of
Inventions that Will Change the World Within Five
Years, New York Times, 17 Dec 2012
http://asmarterplanet.com/blog/2012/12/the-ib
m-5-in-5-our-2012-forecast-of-inventions-that
-will-change-the-world-within-five-years.html

5
Notices ®

This information was developed for products and services offered in the U.S.A.
IBM may not offer the products, services, or features discussed in this
Trademarks
document in other countries. Consult your local IBM representative for
information on the products and services currently available in your area. Any IBM, the IBM logo, and ibm.com are trademarks
reference to an IBM product, program, or service is not intended to state or
imply that only that IBM product, program, or service may be used. Any
or registered trademarks of International
Business Machines Corporation in the United Redbooks ®
functionally equivalent product, program, or service that does not infringe any States, other countries, or both. These and
IBM intellectual property right may be used instead. However, it is the user's other IBM trademarked terms are marked on
responsibility to evaluate and verify the operation of any non-IBM product, their first occurrence in this information with the appropriate symbol (or),
program, or service. indicating US registered or common law trademarks owned by IBM at the time
this information was published. Such trademarks may also be registered or
IBM may have patents or pending patent applications covering subject matter common law trademarks in other countries. A current list of IBM trademarks is
described in this document. The furnishing of this document does not give you available on the Web at http://www.ibm.com/legal/copytrade.shtml
any license to these patents. You can send license inquiries, in writing, to:
IBM Director of Licensing, IBM Corporation, North Castle Drive, Armonk, NY The following terms are trademarks of the International Business Machines
10504-1785 U.S.A. Corporation in the United States, other countries, or both:

The following paragraph does not apply to the United Kingdom or any IBM
other country where such provisions are inconsistent with local law: ILOG
INTERNATIONAL BUSINESS MACHINES CORPORATION PROVIDES THIS InfoSphere
PUBLICATION “AS IS” WITHOUT WARRANTY OF ANY KIND, EITHER Rational
EXPRESS OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF NON-INFRINGEMENT, MERCHANTABILITY OR FITNESS Redbooks®
FOR A PARTICULAR PURPOSE. Some states do not allow disclaimer of Redguide™
express or implied warranties in certain transactions, therefore, this statement Redpaper
may not apply to you. Redbooks (logo)
System z
This information could include technical inaccuracies or typographical errors. Tivoli
Changes are periodically made to the information herein; these changes will be
incorporated in new editions of the publication. IBM may make improvements WebSphere
and/or changes in the product(s) and/or the program(s) described in this
publication at any time without notice. The following terms are trademarks of other companies:

Any references in this information to non-IBM Web sites are provided for Other company, product, or service names may be trademarks or service
convenience only and do not in any manner serve as an endorsement of those marks of others.
Web sites. The materials at those Web sites are not part of the materials for
this IBM product and use of those Web sites is at your own risk.
IBM may use or distribute any of the information you supply in any way it
believes appropriate without incurring any obligation to you.
Information concerning non-IBM products was obtained from the suppliers of
those products, their published announcements or other publicly available
sources. IBM has not tested those products and cannot confirm the accuracy of
performance, compatibility or any other claims related to non-IBM products.
Questions on the capabilities of non-IBM products should be addressed to the
suppliers of those products.
This information contains examples of data and reports used in daily business
operations. To illustrate them as completely as possible, the examples include
the names of individuals, companies, brands, and products. All of these names
are fictitious and any similarity to the names and addresses used by an actual
business enterprise is entirely coincidental.
Any performance data contained herein was determined in a controlled
environment. Therefore, the results obtained in other operating environments
may vary significantly. Some measurements may have been made on
development-level systems and there is no guarantee that these
measurements will be the same on generally available systems. Furthermore,
some measurements may have been estimated through extrapolation. Actual
results may vary. Users of this document should verify the applicable data for
their specific environment.
COPYRIGHT LICENSE:

This information contains sample application programs in source language,

which illustrate programming techniques on various operating platforms. You
may copy, modify, and distribute these sample programs in any form without
payment to IBM, for the purposes of developing, using, marketing or
distributing application programs conforming to the application programming
interface for the operating platform for which the sample programs are written.
These examples have not been thoroughly tested under all conditions. IBM,
therefore, cannot guarantee or imply reliability, serviceability, or function of
these programs.
This document, REDP-4974-00, was created or updated on June 3, 2013.

© Copyright IBM Corp. 2013. 6