5 IBM BUILDING A THOUGHT LEADERSHIP SERIES FOUNDATION TO SUPPORT THE RISE OF COGNITIVE COMPUTING

EMERGING TECHNOLOGIES in COGNITIVE COMPUTING and MACHINE LEARNING

2017 | OCT THOUGHT LEADERSHIP SERIES | October 2017 2

THE NEW ERA OF COGNITIVE COMPUTING AND MACHINE-LEARNING A new era of cognitive computing is unfolding and its impact is already being felt across industries, from preventative maintenance at manufacturing plants and patient diagnosis at hospitals, to the rise of sophisticated chatbots ready to assist us across the connected world. According to estimates, the market around cognitive computing products and services will grow to $46 billion in just a few years and intelligent applications will spread like wildfire transforming our lives at work and at home.

The goal of cognitive computing is straight- tive computing systems must understand con- operating expenses. There were also a pleth- forward: to simulate human thought processes text. They can identify and extract contextual ora of challenges, from harnessing new data in a computerized model. Through a variety elements such meaning, syntax, time, location, sources, to delivering real-time analytics. On of established technologies such data mining, appropriate domain, regulations, user profile, the road ahead, expect to see a lot of cases pattern recognition, and natural language task, and goal to present information that is where IoT and cognitive computing technol- processing, we can build systems that mimic appropriate for an individual or dependent ogies are combined to deliver new insights. how the human brain works. These systems application. The advantage that machine learning offers rely on machine learning and deep learning To fully deliver value, these systems also organizations—the ability to automatically algorithms to automatically learn and adapt need to be able to easily interact with both their build models that can analyze huge volumes from experience without being explicitly pro- human and machine colleagues. For humans, of data and deliver lightning fast results—has grammed. The experience we provide cogni- that typically means speech and text. The rise also led to a growth in the availability of both tive computing systems is data. of cognitive business applications that deliver commercial and open source frameworks, li- Cognitive computing systems need massive a more engaging user experience is a key de- braries and toolkits for engineers. These re- volumes and varieties of data. The more data, velopment in this area, as well as continued sources are democratizing access to machine the more the system can learn, based on previ- advancements in natural language processing. learning for companies big and small and will ous computations, to produce accurate results For machines, the most exciting area is the continue to play a pivotal role in the spread of and predictions. Meeting the scalability and Internet of Things. According to estimates, the machine learning throughout businesses. performance requirements of cognitive com- IoT universe is on track to exceed over 25 bil- Ultimately, cognitive computing, along puting systems involves complex data connec- lion devices by 2020. Right now, there are more with machine learning, is about increas- tions and serious processing power and storage things connected to the Internet than people on ing efficiency, productivity, and innovation to support high-speed data exploration. the planet. From home appliances and cars, to via data analytics and automation. Today, Out of necessity, the future of IT infra- light bulbs and livestock, if you can attach a sen- many people associate them with high-pro- structures at many organizations will become sor to it, it can be become part of a universe of file examples like self-driving cars and ro- increasingly hybrid; where existing operational physical objects able to communicate and inter- bots, which bring the subject good publicity. systems on premise are augmented with new act digitally. This connectivity means more data, However, harnessing value from these tech- data sources being stored and analyzed in the gathered from more places, than ever before. nologies is far from the exclusive domain of cloud. The spike in Apache Spark usage this year A recent survey found that 29% of Big billionaire entrepreneurs and Fortune 500 in the cloud is a great example. Moving forward, Data Quarterly readers already have IoT proj- companies. The growth of cloud services, flexible deployment and integration options ects underway. The top reasons cited for im- lower storage costs, more efficient data pro- will be a must-have for big data projects. plementing IoT projects included increasing cessing options and, open source tools offers In addition to their self-learning capabili- new business revenue sources, increasing cus- resources to organizations of all types and ties, in order to mimic the human brain, cogni- tomer and product knowledge, and reducing sizes to get started today. n THOUGHT LEADERSHIP SERIES | October 2017 3

COOL COMPANIES IN COGNITIVE COMPUTING

No longer the stuff of science fiction, ABBYY provides intelligent capture, intelligently at the edge and “amplify” process the business uses for cognitive computing optical character recognition, innovative intelligence through self-learning, self- and machine learning today include fields language-based, and artificial intelligence assuring business processes for commerce, as diverse as medicine, marketing, defense, technologies to help businesses take action healthcare, and financial services. energy, and agriculture. Enabling these appli- with information. cations is the vast amount of data that com- Crowdflower offers a platform powered by Attivio’s Cognitive Search and Insight panies are collecting from machine sensors, Microsoft Azure Machine Learning that Platform leverages cognitive capabilities, instruments, and websites and the ability to combines machine learning and humans-in- such as machine learning and natural support smarter solutions with faster data the-loop in a single platform for data science language processing to deliver the most processing. relevant information in context but also teams doing sentiment analysis, search It is also clear that we are still in the early offers the flexibility of manual relevancy relevance, or business data classification. days of cognitive computing and machine tuning to optimize results. learning, and to be sure, there are technical, Darktrace was founded in Cambridge, U.K., political, and ethical considerations to be C3 IoT offers a comprehensive technology in 2013 by mathematicians and machine dealt with before this new wave of solutions stack for the rapid design, development, learning specialists from the University of comes closer to reaching its potential. How- deployment, and operation of next- Cambridge, together with world-leading ever, innovative companies do have products generation IoT applications that unlock intelligence experts from MI5 and GCHQ, and services today to help customers put data-driven insights and transform business to bring transformative technology to the more data to work. processes. challenge of cybersecurity. To help readers gain a greater understand- ing about this emerging area of information Cloudera offers a modern platform for Databricks, the company founded by the technology, the solutions available, and their machine learning and advanced analytics creators of the Apache Spark project, recently role in handling real-world challenges, built on open source technologies, and introduced Deep Learning Pipelines, a library DBTA presents the inaugural list of Cool recently introduced its Data Science to integrate and scale out deep learning in Companies in Cognitive Computing. Workbench, based on the company’s Apache Spark, which has the potential to acquisition of data science startup Sense.io, —Joyce Wells accomplish for deep learning what Spark did to accelerate data science and machine for big data—make it approachable to learning for the enterprise. a much broader audience. Alpine Data enables organizations to create Cogitai is dedicated to building artificial Dataiku, whose name is a portmanteau of a culture of analytics at scale by providing intelligences (AIs) that learn continually from data and haiku, espouses the view that data an integrated analytics platform that brings interaction with the real world, with the goal projects should have “a structured process, a machine learning, data, and people together to of building the brains, i.e., the continual- single flow, from start to finish,” and provides create operational solutions for business users. learning AI software that enables everyday the Data Science Studio (DSS), to enable things to get smarter with experience. scalable data science to any organization. Amazon AI services bring natural language understanding, automatic speech recognition, CognitiveScale offers an augmented DataRobot uses advanced enterprise visual search and image recognition, text-to- intelligence platform and the ENGAGE machine learning automation to enable speech, and machine learning technologies and AMPLIFY products that pair humans users to quickly build and deploy highly within the reach of every developer. and machines so they can “engage” users accurate machine learning models. THOUGHT LEADERSHIP SERIES | October 2017 4

DatumBox offers services available through and triaging the most relevant attacks, and SAP has relaunched SAP Leonardo a REST API, including a large number of allowing security analysts to focus their (previously known as its brand for IoT) as off-the-shelf classifiers and natural language resources on only the most dangerous threats. its digital brand, including IoT solutions, processing services which can be used in MapR provides the distributed deep learning to take advantage of advances in big applications for sentiment analysis, topic Quick Start Solution comprised of a data layer, data and analytics, the ability to connect classification, language detection, subjectivity which is managed by the MapR File System people, things, and business with the SAP analysis, spam detection, reading assessment, (MapR-FS) service, a middle orchestration Cloud Platform, and technologies such keyword and text extraction, and more. layer using Kubernetes to manage the GPU/ as machine learning to enable IoT and Digital Reasoning software assembles an CPU resources, and a top application layer Industry 4.0 strategies across digital logistics, integrated circuit of algorithms that organize using TensorFlow as the deep learning tool. manufacturing, and asset management. information into a graph-based knowledge Nara Logics builds a synaptic network of SAS has been solving customer problems model to enable predictions based on a high explicit and inferred connections to create with cognitive capabilities for years, fidelity representation of context, and, like an intelligence layer on top of chaotic, siloed and contributes critical components to the human mind, learns new things, adapts, enterprise data for real-time, context relevant the cognitive computing mix, including and gets smarter over time. recommendations. natural language processing and open, deep H20.ai provides an open source AI OpenText, a provider of enterprise learning API (application programming platform that is used by over 100,000 information management solutions, recently interface) libraries sitting on top of data scientists and more than 10,000 introduced a new artificial intelligence advanced analytics, including the new SAS organizations around the world. platform that combines open source machine Viya platform. IBM introduced Watson to the world in learning to offer users machine-assisted Search Technologies, recently acquired by 2011 when it won the $1 million first place decision making, automation, and business prize on Jeopardy, and, today, the cognitive optimization, in an easy to use package. Accenture, offers a proprietary Content Processing Framework and collection of computing platform encompasses products Progress, a provider of application API-level data connectors—which enable and APIs that bring the power of AI to development and deployment technologies, organizations in more than 45 countries takes a cognitive-first approach, and recently access to unstructured enterprise data across and across 20 different industries to acquired DataRPM, a provider of cognitive disparate and legacy systems—that will uncover business-critical insights. predictive maintenance software for the be integrated into the Accenture Insights Platform, helping clients embed analytics Infosys offers the Nia platform, which industrial IoT (IIoT) market, to fuel this combines the big data/analytics, ML, KM, and strategy. and AI into their business to generate new intelligence at speed and scale. cognitive automation capabilities of its first- RapidMiner offers a unified data science generation AI platform, Mana; the robotic platform that accelerates the creation of Sinequa provides a cognitive search and process automation capabilities of AssistEdge; complete analytical workflows from data analytics platform for Fortune Global 2000 and the machine-learning capabilities of prep to machine learning to deployment in companies and government agencies. Using Skytree; with optical character recognition, a single environment, improving efficiency advanced natural language processing and natural language processing capabilities, and and shortening the time to value for data machine learning algorithms, the solution infrastructure management services. science projects. offers insights extracted from structured IPsoft, which automates enterprise IT RAVN offers Applied Cognitive Engine and unstructured data. and business processes through the use of (ACE), which offers capabilities that go far SparkCognition, launched in 2014 in digital labor, offers Amelia, an AI platform beyond merely searching for documents and Austin, Texas, provides clients with patented that connects conversations to data and webpages, including machine learning tools AI-powered products to enhance cyber processes in order to enable personalized and automatic extraction of key data from security, and state-of-the-art machine service to customers. the unstructured data under management. learning technology to identify and prevent JASK, an enterprise artificial intelligence (AI) Saffron Technology offers patented equipment failures before they happen. cybersecurity company, that recently launched technology that mimics humans’ natural with the announcement of $12 million in ability to learn, remember, and reason Tamr offers patented software that fuses the Series A funding, offers a cloud platform in real-time and is targeted at defense, power of machine learning with knowledge that uses machine learning and AI to deliver healthcare, manufacturing, and financial of an organization’s data to automate the end-to-end network monitoring—identifying services. rapid unification of data silos at scale. n sponsored content THOUGHT LEADERSHIP SERIES | October 2017 5

Building a Foundation to Support the Rise of Cognitive Computing

We are at the start of the cognitive era. called the “data corpus” and according to BUILDING THE RIGHT HYBRID DATA The potential advantages spoken about for so IBV could include a variety of data types MANAGEMENT SYSTEM AS YOUR long are within reach and many businesses such as structured and unstructured data COGNITIVE BUSINESS’S FOUNDATION are realizing that now is the time to act. housed in multiple locations5 whether that At first glance, it may be difficult to According to a recent study from IBM’s happens to be a database, data warehouse, determine what you can do to get ready for Institute for Business Value (IBV), 73% of Hadoop or other open source or proprietary the cognitive era. We recommend that you the CEOs that it surveyed across the globe repositories. Some of this information may establish a solid foundation by building out thought that the role cognitive computing even come from previously untapped sources a data management ecosystem that delivers would play in their company in the future that can now add value.6 the flexibility and performance required by would be important.1 These CEOs aren’t Despite the fact that the corpus is specific cognitive solutions. referring to a distant future that might be to the needs of the cognitive solution, a realized ten or fifteen years from now, 50% proper data foundation should be put in Choose a solution with the flexibility indicated a 2019 timeframe for their own place ahead of time across the organization. necessary to provide the data you need cognitive computing adoption.2 If you’ve Having better data quality, more data, and The corpus of a cognitive solution been thinking about implementing cognitive easier access to that data will pay off when will need to draw upon structured, technology, now is the time to act. being designated for cognitive solutions. unstructured, and semi-structured Businesses are excited about the near- Early focus in this area will have a profound information from a multitude of sources term prospects of cognitive computing impact on the knowledge that can be gleaned and data management systems, so make for good reason. By taking advantage of when making data-driven decisions. sure querying data from all those locations cognitive computing’s ability to support can be done easily. Technologies that and build upon human expertise, both the allow you to write queries once and run quality and consistency of decision making ENSURING ACCESS TO ANALYTICS them anywhere, like those found in IBM’s can be improved. This is an opportunity for Analytics similarly plays a large role for family of data management solutions, are businesses to drive significant measurable cognitive solutions by delivering the insights a good starting point. If you want to run value. In a recent survey, executives indicated which make such solutions helpful. While queries on-premises today and in the cloud anticipating a return on investment of 15% descriptive, predictive, and prescriptive tomorrow, or vice-versa, you won’t need to for cognitive initiatives.3 analytics can all play a part, data science and spend time rewriting them. This means you In order to take advantage of the cognitive machine learning have received considerable can run your data warehouse the way you future, we must look at the core foundational attention as of late for being an achievable want without location-based impediments. components that must first be put in place. early step toward a more complete cognitive Additional data source flexibility can In its definition of a cognitive business, strategy. In fact, 44% of the organizations that be provided through techniques such as the IBV mentions two key components: IBV identified as “cognitive capable” in its data virtualization, which allow you to an ability to use a range of analytics study had adopted machine learning already.7 conveniently bring in external data for capabilities such as descriptive, predictive Unsurprisingly, a healthy data ecosystem use in analytics. Since not all data capable and prescriptive analytics as well as a has a role to play here as well. Data of delivering insight lives within your data management ecosystem that handles management solutions that are flexible organization, being able to draw upon that multiple sources and a high volume of data enough to draw upon multiple data sources information is crucial. Yet, even data inside (both structured and unstructured).4 and integrate well with various analytics the organization will see benefits from data tools are necessary to deliver the most virtualization. Data virtualization can help comprehensive insights. Not only does you connect to Hadoop and Spark and will BUILDING A CORPUS OF INFORMATION flexible data management allow you to make it easier to access data or repositories For a cognitive solution to reliably add prepare for the cognitive future, but also that might have been moved. value, the solution must have a substantial benefits the analytics you are currently Of course, it is also helpful if your amount of information upon which to draw. running with smoother interactions and licensing options are just as flexible. If you This collection of information is typically depth of referenceable data. believe your business might move toward sponsored content THOUGHT LEADERSHIP SERIES | October 2017 6

having more cloud-based data management like BLU Acceleration® in IBM’s data Similarly, pre-tuned unified data technology concurrent with, or as a result management products, help to scale out platforms that contain both a data warehouse of, the push for cognitive solutions, it’s capabilities. In turn, this can significantly and data science tools can help address the worth investigating options that will let you improve performance by ensuring that you need for access to analytics within the data switch on-premises for on-cloud licenses as can take advantage of the increasing amount management ecosystem. The data science the need arises. This will allow you to switch of data available for your cognitive needs. needed for the cognitive era can be done your company’s technology landscape To mitigate rising storage cost, look faster when machine learning is run in the exactly when the time is right. for solutions that provide both actionable same place as the data. Yet, even if you choose compression, which preserves the order not to purchase a platform such as that, Choose a solution that can deliver the of data so that it can be used for analytics integration with other analytics software performance and efficiency needed for the without decompression, as well as data is vital. Make sure your data management cognitive era skipping so that unnecessary data processing solutions have the capability to work with Still, flexibility is meaningless if a solution can be bypassed. Features such as these save and allow you to take advantage of the doesn’t have the power to back it up. High both time and money. multitude of analytics options available. performance from your data management Quick time-to-insight is also critical for architecture is essential to meet the stringent cognitive solutions and is dependent upon demands of cognitive computing, because the speed of analytics and speed of setup. The WHERE TO START LOOKING insights need to be delivered within a timely speed of analytics for cognitive technologies With the cognitive era close at hand, now manner to add the most valuable. The needs to be as close to real-time as possible is the time to start preparing, particularly if massive volume of data, speed of uncovering so that insights can be delivered the instant your CEO was one of the 50% surveyed who insights, and access to analytics required they’re needed. Again, look for solutions with has plans for cognitive adoption by 2019.8But must be met or the effectiveness of the in-memory columnar technologies that are you don’t have to face that daunting task cognitive solution will suffer. capable of providing insight from real-time alone. Get started in the right direction by The increasing amount of data being operational and historical data. In addition, learning more about a comprehensive family collected requires that a data management consider how quickly the solution needs to be of data management products from IBM ecosystem be both scalable and efficient. One up and running. Cloud offerings may provide which has the flexibility and performance way to achieve this is through in-memory the quick setup you need, but a pre-tuned, on- that cognitive solutions will require. As the columnar technologies that take advantage premises option might be better depending on cognitive era transitions from theoretical to of massively parallel processing (MPP) based your situation. Work with a vendor like IBM reality, begin laying the foundation that will cluster architectures. These technologies, that is capable of providing both. empower your business. n

1 Abercrombie, Cortnie, Rafi Ezry, Brian Goehring, Anthony Marshall, and Hiroyuki Nakayama. “Accelerating enterprise reinvention” IBM Institute for Business Value. June 2017. https://www-935.ibm.com/services/us/gbs/thoughtleadership/accelentreinvent/ 2 Abercrombie, Cortnie, Rafi Ezry, Brian Goehring, Anthony Marshall, and Hiroyuki Nakayama. “Accelerating enterprise reinvention” IBM Institute for Business Value. June 2017. https://www-935.ibm.com/services/us/gbs/thoughtleadership/accelentreinvent/ 3 Abercrombie, Cortnie, Rafi Ezry, Brian Goehring, Anthony Marshall, and Hiroyuki Nakayama. “Accelerating enterprise reinvention” IBM Institute for Business Value. June 2017. https://www-935.ibm.com/services/us/gbs/thoughtleadership/accelentreinvent/ 4 Ezry, Rafael, Michael Haydock, Bruce Tyler and Rebecca Shockley. “Analytics: Dawn of the Cognitive Era.” IBM Institute for Business Value. October 2016. http://www.ibm.com/business/value/2016analytics/ 5 Sarkar, Sandipan and David Zaharchuk. “Your cognitive future – How next-gen computing changes the way we live and work, Part II: Kick-starting your cognitive journey.” IBM Institute for Business Value. March 2015. http://www.ibm.com/business/value/cognitivefuture 6 Sarkar, Sandipan and David Zaharchuk. “Your cognitive future – How next-gen computing changes the way we live and work, Part II: Kick-starting your cognitive journey.” IBM Institute for Business Value. March 2015. http://www.ibm.com/business/value/cognitivefuture 7 Ezry, Rafael, Michael Haydock, Bruce Tyler and Rebecca Shockley. “Analytics: Dawn of the Cognitive Era.” IBM Institute for Business Value. October 2016. http://www.ibm.com/business/value/2016analytics/ 8 Abercrombie, Cortnie, Rafi Ezry, Brian Goehring, Anthony Marshall, and Hiroyuki Nakayama. “Accelerating enterprise reinvention” IBM Institute for Business Value. June 2017. https://www-935.ibm.com/services/us/gbs/thoughtleadership/accelentreinvent/