1. best possible solution to a given problem,

1. IntroductionOver the last century, we went from numbers (1900’s) to data (1950’s) to knowledge (2010’s) 1. It’s not about replacing one type with another, but rather enhancing it. We were able to create data by processing numbers using programmable systems whereas cognitive computing allows us to make sense of the data. The tabulation systems were all about adding objects to the inventory, the programmable systems were about taking the small details like color and matter into consideration; and cognitive systems are about predicting the weather forecast. For decades now, computers have been swift at processing and calculations than humans, but they haven’t been able to accomplish simple tasks like understanding the natural language or recognizing distinct objects within an image. The enormous volume of redundant data that is being generated across the globe is creating a hamper for customers and professionals; thus, we need systems that would excel at extracting and making data useful 2.Cognitive Computing is the one which uses machine learning and natural language processing to empower interaction between people and machines more naturally to amplify cognition and human expertise. A system that would provide information to the end-users to help them decide the best possible solution to a given problem, as human experience and judgment play a crucial role. The main intention of cognitive computing is to mimic the human cogitation process using self-learning algorithms as a computerized model. The computer system can simulate the way the human brain functions 3.The Turing test 4 or the ability of a computer to imitate humans is not the only measure of success of cognitive computing rather it will be measured in more practical ways, like the number of live saved, diseases cure and new business venture opportunities.2. BackgroundWhen IBM’s Watson defeated human opponents (Brad Rutter and Ken Jennings) in the Jeopardy Challenge in 2011 5, we knew that a new kind of technology was emerging. This was the beginning of cognitive computing and the end of the so-called AI winter. All the systems that were designed until now used to give us information that was allocated previously in the memory (if the data was put in the right place). The questions asked each time was also something the system expected. But, IBM Watson did things a bit differently, here a huge amount of unstructured information was made accessible always instead of storing it in a single location. With cognitive systems becoming mainstream in today’s world a lot of jobs are at risk. The most important thing to remember is that Watson cannot work without a cognitive background and will always need the assistance of humans to train him, to be more relevant 6. Once the system is mature enough there would be drastic changes in the way we do computing. We wouldn’t use computers as just “regular computers” rather we would interact with it as assistants that will provide valuable information instead of raw data. Watson has been able to conquer complex data sets and develop a niche skill for learning, understanding and reasoning way beyond just deciphering 5. It was built to combine data analytic and reasoning with common sense, ethical values and most importantly human attributes. The team at IBM created a system that was designed to adapt and mine useful information from unstructured data. These systems can read the text, see the images and hear the voices, but more than anything they are able to interpret information, organize it and give a valid explanation of what it represents.Watson encompasses these 5 core capabilities which makes this cognitive system unique from others in the market 5:• Create deeper human interaction – Take the data that is available in today’s world and create pictures with minute details like sentiments, emotions, environmental state, tone and nature of a person’s relationships to enhance human interactions with each other.• Elevate the level of expertise – Knowledge and technology is growing and changing beyond the rate that we can keep up with. The immense amount of resources available reduces the time for a professional to become an expert.• Instill services and products with cognition – Wherever code and data go, cognition will always follow. Nobody ever imagined the unceasing improvement and adaptability of augmentation.• Amplify exploration and discovery – The cognitive computing platform must be vast and pliable enough to be applied by organization to any industry worldwide since the most powerful tool a business can possess is innovation.• Equip cognitive processes and operations – Businesses with cognitive processes provide awareness about constant learning, workflow model, efficient results. The key factor of incorporating cognitive mode of operation is to enhance the decision-making ability to meet the speed of data being generated each day.Due to Watson’s rapid growth we are getting a sense of what’s possible and what’s achievable. An example of how Watson is really making a difference in today’s world: Oil and gas companies can merge seismic imaging data along with thousands of reports, papers, articles, weather forecasts to create danger and safe scenarios for exploratory drilling or examine the attendance, scores and behavior of a student to create personalized education plans and for the school to maintain accurate student records 5.3. Analysis3.1 Architecture of IBM WatsonWatson is an IBM supercomputer that incorporates artificial intelligence (AI), cognitive computing and sophisticated analytical software as an optimal “question answering” machine 7. Here are the hardware specifications of IBM Watson 7:• 90 IBM Power 750 Servers – each uses 3.5 GHz POWER7 8-core processor, 4 threads per core, 2880 simultaneous threads • 16 TB of RAM• 80 Tera FLOPS• 500 GB/sec or 1 million books/sec• Each Power 750 Server costs about $34500• The entire system is worth about $3 millionThe POWER7 multiprocessor consists of the following properties 8:• It is a superscalar symmetric multiprocessor (especially developed for supercomputers) with a high-performance VLSI chip which has eight quad-threaded cores• 256 CPU’s of hardware coherency fabric• Contains 1.2 billion transistors and is fabricated on 45nm CMOS Silicon-on-Insulator technology with 11 levels of low-k copper wiring• Consists of L1 cache (32 KB/core), L2 cache (256 KB/core), L3 cache (32 MB which accessible by all the other cores and an embedded DRAM)• To achieve high efficiency uses the truculent out of order instruction execution using the available paths Figure 3.1: IBM POWER7 CPU die• Includes simultaneous multithreading of 3 types (1, 2, 4-way simultaneous multithreading) and intelligent threading• Has 2 modes of flexible processor packaging (TurboCore and MaxCore mode)• 2 DDR3 memory controllers which consists of 4 channels each, operates at 6.4 GHz and can address up to 32 GB of memory Figure 3.2: Multithreading EvolutionHere are the software specifications of IBM Watson 7:• IBM’s DeepQA 9• Apache UIMA (Unstructured Information Management)• SUSE Linux Enterprise Server• Apache Hadoop3.2       How IBM Watson understands language? Figure 3.3: DeepQA high level architecture depicting the different stages of how IBM Watson understands language 10• When Watson is presented with a question, it parses the question and extracts valuable information from the question.• Then, a set of hypotheses is generated using various sources that have some potential for carrying a valuable response.• Next, a deep comparison between the question and potential response using various reasoning algorithms is performed. This step is the most challenging one. Based on the criteria chosen to compare, there exists various reasoning algorithms and the best one is chosen by Watson.• Based on the selected subject, each of the reasoning algorithm provides a score which denotes the extent to which the response is relevant to the question.                                                             Figure 3.4: Prediction versus percentage attempted 11• This score is then compared against a statistical model which tell us how well the algorithm established an inference between all the possible responses. The statistical model depicts the level of confidence that IBM Watson has about the evidence from which candidate answer is chosen.Watson’s level of confidence can be summarized using the model.                                 Figure 3.5: Baseline performance 11• This process is repeated for each of the responses until Watson finds a response that is more strong and relevant than the others.For instance: Here we see a comparison of two candidate answers produced by the system to choose between Argentina and Bolivia. Search engine prefers Bolivia due to popular border dispute whereas Watson performs Argentina (the correct answer). Based on the various criteria such as Location, Passage support, Popularity, Source Reliability and Taxonomic a comparison of two candidate answers have been made and depicted in the graph below.                                   Figure 3.6: Evidence profile for two candidate answers 113.3      How Cognitive Computing is Incorporated into Applications            IBM has built Watson Developer Cloud to help companies around the globe provide an open cognitive platform which is secure and gives easy access to the API’s 12. These interfaces assist in incorporating cognitive into various kinds of applications, operations and products. The API’s can be classified as follows:• Language: Classify the natural language simple text and long conversations, translate the language, retrieve the passage, extract the relationships, sentiment analysis.• Vision – Excerpt meaningful information from images and derive significant value.• Data Insights – Find out the most targeted search and derive a trend analysis• Speech – Used to convert speech to text and vice versa and the ability to train using custom language models.          A combination of the above API’s can be used to solve numerous analytical business problems and create an enriching experience for the user. Also, the amalgamation of data analytics with cognitive services helps in complex discoveries and carry out decisions driven by insight. Figure 3.7: Process of incorporating cognitive capabilities 12The Watson Developer Cloud services needs to be merged with the existing services so that cognition can be enabled. Some of the key components of the IBM Cloud are:• Corpus: Logs, video and audio content, trouble tickets are all ingested by the cognitive services. They can be stored in various repositories but will be highlighted as a single block to show the importance of corpus.• Watson Framework and API’s: These interact with all the other components. It also includes Watson Explorer which lets the cognitive computing understand and analyze unstructured documents. • User Interface (UI) – A set of visual menus and interfaces that give the user quick access to data to simplify the editing process, this is a dashboard. We can build the new dashboard over existing ones and can interact with other cognitive services.• Orchestration Engine: This synchronizes cognitive services with other services and coordinates the sequencing of steps and activities based on the rules of the business model. It can either be built from scratch or on an existing engine.  • Existing Systems: The systems in an enterprise which need cognitive computing capability to be embedded in them fall in this category. Another impressive feature of Watson Question and Answer API 13 is that it allows various applications to interact with Watson. With this API, we can ask questions to Watson, get the response and give the feedback on those responses. Each of the responses that Watson produces has a supporting evidence of some form i.e. passages, articles, documents. This API could be embedded into an application in an ample number of ways based on the business need. Watson provides two different ways of using the Question and Answer API:• Asynchronous Mode: When a question is thrown at Watson, the reply it returns is a link to retrieve the answer when it is ready. To check the whether the question has been completely processed, a server is used and to make sure that the response contains the final answer, a status field is used.• Synchronous Mode: Here, we don’t throw the question directly at Watson, instead we do it through a POST operation. This feature allows us to retrieve the answer to the question in a synchronous manner. Also, the need for a server and status field is eliminated because of the POST operation.It is of prime importance to keep in mind the nature of the cognitive platform to ensure that the new system can be built effectively and efficiently. We should also be able to run ample number of test and use cases to check whether it meets the client’s requirements 12. 3.4      Limitations, Constraints and Expected Enhancements            From a technical point of view, cognitive computing was designed to make use of vast amount of data. The systems needed huge quantities of raw material from which it would extract valuable information, learn and predict. Large global organizations 14 had access to this kind of data whereas the small and medium sized organization would struggle to gather more data. Another challenge is training the cognitive systems. To train the cognitive system, not only sufficient data required but also skilled resources who can invest time in tuning the engine to ensure valuable outputs can be obtained. This maybe a barrier for new organizations entering the industry. The next challenge is, to harness the full potential of cognitive computing cost plays an important role and the investors would like to keep the rate of interest in mind before starting such new ventures. Furthermore, significant legal and privacy associations are there to fetch the data, especially when people perceive some information as personal such as emails, search history and downloads. Laws vary from country to country and business rules constantly keep changing.Keeping Watson in mind, one of its greatest revelation is that while using Watson to help answer the question, a person may realize that he might have asked the wrong question 10. In the process, the machine helps you ask questions in a different manner, ask better question and more important questions to reconsider your problem in a different aspect. It might open doors for people to understand the competitive threats and new opportunities in the market that one would have never thought about.This is probably one of the most promising applications of cognitive computing, IBM is collaborating with a lot of leading cancer institutes to enhance the clinical trials and provide personalized treatment plans for patients around the world 5. The algorithm running inside Watson can reduce the time from weeks to minutes to translate DNA insights and evaluate a patient’s genetic profile. This result allows the doctor to concentrate on the specific cancer-causing genetic mutation. IBM is working on expanding the use of Watson to enable global collaboration and innovation. with the collaboration of IBM and SoftBank, expansion of Watson’s ability to communicate and “Think” in various languages, there are bound to be further technological improvements. This merger would enable us to bring cognitive computing to every nook and corner of the world.80% of all the data that is available today is dark data i.e. we can’t make sense of that data and the amount is expected to grow to over 93% by 2020 15. Hence, cognitive is trying to bridge the gap between the dark data and useful data to help industries sustain in the market.• Oil and Gas – Cognitive computing can help companies prevent drilling in incorrect places and with flow optimization.• Retail – Data from various social networking websites can help us identify buying pattern, insights, preferences and what are the changes in the trends in the society.• IoT (Internet of Things) – NLP (Natural language processing) is unmatched in Watson and with rise in machine-to-machine data (noisy unstructured data), it is perfect for implementing cognitive computing.• Security – It is no longer about viruses, threats or firewalls but rather about behavioral analysis of people and systems in real time.• Energy – The digital meters across the globe produce dark data and it is difficult to incorporate renewables without understanding the current market demands.• Healthcare – It’s a huge industry known for disruption and new forms of insight 16.• Transportation – By 2020, three-quarters of the cars in the world will be connected. They would be needed to make real-time decisions about the driver’s behavior and the environment.      There’s an enormous scope for cognitive computing to play a key role in improving the way the industry functions and helping change people’s lives. This is as good as it can get.4. ConclusionThe Cognitive Era is the next big thing in the field science and technology to analyze and ameliorate the human condition. Technology opens doors to a world of possibilities and opportunities, but the future depends on the choices we make. Cognitive computing will possibly change the nature of work done by people and help us perform tasks faster and more accurately. Processes will be cheaper and more efficient. Just as how cognitive systems are the machines inspired by the human brain, they will also inspire the human brain to reason and assess the way in which we think. In today’s world it’s not about knowing all the answers rather the ability to ask efficient and effective questions, marks a true genius. The IBM Watson supercomputer was no joke, the scientists who designed this beast spent years on researching about artificial intelligence and natural language processing to produce a series of breakthroughs. It is used to help doctors diagnose diseases and asses the best treatment for patients. The main intention of Watson isn’t to replace doctors but act as a useful assistant to the doctor 17. In this paper, we were able to reason out the need for cognitive computing and how it came into existence. Looking at the in-depth architecture of IBM Watson supercomputer and how cognitive computing is incorporated into this machine just showed us how powerful of a beast it is. The limitations, constraints and the expected enhancements gave us an insight of what’s not possible and what to expect in the upcoming days. The rapid evolution of IBM Watson has opened our eyes to all the different things that’s possible in the field of cognitive computing.