Computing machines are capable of holding an ocean of knowledge
and also can recall relevant material useful to resolve dilemmas for competing
in sports or provide expert advice for evidence based medicine at lightning
speed. Machines have learned to sense, respond, debate, discuss, and interact
with reasoning to achieve such extraordinary capabilities which have been the
territory of only expert human beings so far.
The preceding statements are reflective of the vast variety
and broad spectrum of applications with varied maturity levels and much
promise. There are a few points to note when we come across Machine learning
and Big Data connecting them to Artificial intelligence.
1. There is much to unlearn about information systems built
when data, information storage, and access were at a premium.
It was necessary for engineers to follow computing models
that required instructions for anticipating every possible scenario to develop
information systems. Engineers were limited by programmable computing (a.k.a Von Neumann
computing) to build systems that helped people both at work and in their
personal lives. They still prevail for solving numerous business problems.
In the age of big data, the nature of computing is changing to fit the needs of more complex and uncertain scenarios. Computing has now begun to show the potential to adapt to humans naturally in every walk of life.
In the age of big data, the nature of computing is changing to fit the needs of more complex and uncertain scenarios. Computing has now begun to show the potential to adapt to humans naturally in every walk of life.
2. Nature of decision making and problem-solving systems of
the industrial age emphasized the left brain thinking based on logic and
analysis for the most part.
The left brain thinking became the foundation of the
automation mechanisms built to aid human and organizational endeavors. Left
brain thinking and the infrastructure in the analytics space encompass
applications built around traditional databases. These applications evolved and
contribute to Business intelligence, broadly covering data warehousing, online
analytical processing, and meta-data repositories. While all of them use past
data and experience for knowledge and discovery – there are differences. The
exercise is beyond the scope of this notes. The one thing in common is that
they use sorting and searching algorithms with sequences of instructions which
lead to responses from the systems.
3. The potential of Big data and Machine learning is
enabling to explore and build right brain systems which are taking shape as the
AI ecosystem.
Smart, intelligent, cognitive systems support the Artificial
intelligence (AI) ecosystem. Cognitive computing systems are developing human capabilities
and interactions, and scale better than ever before. Machine learning, prescriptive and predictive
analytics, deep learning, recommendation engines, evidence-based expertise,
narrative science and much more comprise the Artificial intelligence (AI)
ecosystem backed by Cognitive computing.
In the consumer and business space, the technologies of
analytics, big data, cloud, and AI are converging to form the AI ecosystem. The
AI systems are capable of deriving inferences and pursue the objectives with
the data provided. The AI systems draw on the fundamentals of statistics and
data mining algorithms with iterative, improved learning from real time
data.
4. The AI ecosystem as heard in the business communities represents
a higher maturity level of analytics compared to the descriptive, predictive,
and prescriptive analytics practices.
Examining the types of questions asked and problems solved is a way to probe into the state of capabilities of the systems for the business. Tom Davenport brought up analytics maturity levels at the organization level more than a decade ago on how to take the first steps for exploring data based decision making in his work “Competing on Analytics.” He later discussed the three stages of Analytics maturity (Davenport, T. H. (2013)). He called the era of Business Intelligence as Analytics 1.0 (up to the mid-2000s with enterprise data warehousing capabilities and business intelligence software), Analytics 2.0 as the era of Big data (internet based social media and networks begin to amass data), Analytics 3.0 as the era of building analytical power into consumer products.
Examining the types of questions asked and problems solved is a way to probe into the state of capabilities of the systems for the business. Tom Davenport brought up analytics maturity levels at the organization level more than a decade ago on how to take the first steps for exploring data based decision making in his work “Competing on Analytics.” He later discussed the three stages of Analytics maturity (Davenport, T. H. (2013)). He called the era of Business Intelligence as Analytics 1.0 (up to the mid-2000s with enterprise data warehousing capabilities and business intelligence software), Analytics 2.0 as the era of Big data (internet based social media and networks begin to amass data), Analytics 3.0 as the era of building analytical power into consumer products.
5. Machine learning is different. The talk about machine
learning is around predictive or descriptive learning using big data applying
statistical and mathematical models that scale.
Data mining models and applications come closest – except they use large data bases. Typically, data mining tasks are grouping people or objects based on selected characteristics, estimating for inferring or guessing reasonably, predicting expecting behaviors as a consequence of previous actions. The list goes on with more statistical models with deeper selective techniques for evaluating relationships and associations.
Data mining models and applications come closest – except they use large data bases. Typically, data mining tasks are grouping people or objects based on selected characteristics, estimating for inferring or guessing reasonably, predicting expecting behaviors as a consequence of previous actions. The list goes on with more statistical models with deeper selective techniques for evaluating relationships and associations.
The ability to question determines the level of
understanding to find solutions. With descriptive analytics, there is an
understanding of the power of analytics for starters. Progressing to predictive
and prescriptive analytics lets the business come up with frameworks to arrive
at desired results experimenting with product mixes. Machine learning overlaps
with the foundations allowing to scale and iterative collaboration with man and
machine. The use cases of predictive analytics are varied – for example,
figuring consumer behavioral patterns, forecasting sales, predicting failures
in machinery or fraud. [Ref: Chapter 6, Hurwitz, J., Kaufman, M., & Bowles, A. (2015)].
Machine learning broadly encompasses supervised learning
(through association, regression, classification) and unsupervised learning for
pattern recognition – an engineering term for classifying. A super market chain
or Amazon would prefer to know the customer’s next choice of items based on
previous purchases. Customer too prefers to check out the cart faster. These
recommendation engines and matching systems work with recognizing behavioral
patterns obtained from the data. Recognizing patterns helps derive a response
of likelihood of a behavior. Detecting frauds, stock market predictions with
changing data and environment require learning and intelligence more than just
depending on the system designed using a large database with traditional,
structured queries. Applications for
recognizing speech and faces, robotics, and biometrics use machine learning
techniques. High-risk scenarios such as
patient health management are yet another use case for advanced analytics
combined with AI ecosystem.
6. A broad spectrum of business analytics applications is
beginning to show up as machine learning applications.
Advanced analytics including data warehousing applications
scenarios are on one side of the spectrum. One such category of questions is
those that predict re-order levels for inventory management based on sales next
quarter. Machine learning is helping when answering questions such as alerting
city security when a large-scale crime scene is unfolding in real time or even
predict one. Applications based on machine learning and a mix of advanced analytics
get better iteration after iteration – as they learn from the previous
happenings. On the other end of the spectrum are the applications that mimic
the brain with intelligence that far surpasses known human expertise in medical
or space sciences. The infrastructure is built on the principles of Neuro Science
and Nano Technology to design cognitive chips to take up part of the tasks the human
brains can accomplish (see Dharmendra Modha and SYNAPSE).
7. Watson, AlphaGo.
Any briefing about AI systems is not complete without mentioning the AI Super computers IBM’s Watson and Google DeepMind’s AlphaGo. There are more in the race – most heard are Microsoft’s Oxford and Baidu’s Minwa. The problems that Watson takes up have come a long way since the initial Jeopardy win in 2011. Watson is being trained to aid doctors in the area of evidence based medicine for treating diseases better. AlphaGo is no longer limited to winning the Go board game. They are image recognition engines with natural language processing powers participating actively in the AI ecosystem.
Any briefing about AI systems is not complete without mentioning the AI Super computers IBM’s Watson and Google DeepMind’s AlphaGo. There are more in the race – most heard are Microsoft’s Oxford and Baidu’s Minwa. The problems that Watson takes up have come a long way since the initial Jeopardy win in 2011. Watson is being trained to aid doctors in the area of evidence based medicine for treating diseases better. AlphaGo is no longer limited to winning the Go board game. They are image recognition engines with natural language processing powers participating actively in the AI ecosystem.
Conclusions:
The primary message is to focus on the ability to compute
differently and use data uniquely to look beyond the horizons seeking the
unexplored and understand subtleties. The structured and unstructured data from
various sources is beginning to come out of the dark to shed light on the
patterns, correlations to match data with deep expertise to answer tough
questions for businesses. The broad and vast maturity levels of the analytics
and big data capabilities is the guiding force for enabling AI ecosystem
growth. To take the first step is to set the stage for a vision towards the
organization’s power to tap into the potential of machine learning and big
data. Cognitive systems with the power of big data and machine learning are to
become the sources of new dynamic knowledge sources to think and act
differently on information for businesses.
References:
Alpaydin, Ethem. Introduction to Machine Learning, edited by Ethem
Alpaydin, MIT Press, 2014.
Davenport, T. H. (2013). Analytics 3.0. Harvard Business Review, 91(12), 64-72.
Kelly, III, John E., and
Steve Hamm. Smart Machines: IBM's Watson and the Era of Cognitive Computing,
Columbia University Press, 2013.
Hurwitz, J., Kaufman, M., & Bowles, A. (2015). Cognitive computing and big data analytics
Hurwitz, J., Kaufman, M., & Bowles, A. (2015). Cognitive computing and big data analytics
Loshin, David. Business Intelligence, The Savvy Manager’s
guide, Morgan Kauffman Publishers, 2003.
August
2011 Communications of the
ACM: Volume 54 Issue 8, August 2011
What to expect from Artificial intelligence, Sloan
Management review, Spring 2017.