Data Democratization Helps Artificial Intelligence and Machine Learning to Produce Better Results

Emerging technologies like Artificial Intelligence (AI) and Machine Learning (ML) have initiated a Ripple Effect to bring about massive industrial revolution, touching almost every work area and impacting almost every business whether big or small, so much so that the technologies by themselves cannot be ignored anymore. If you are well connected with my research work and regularly read my posts, you must be having a fair idea that these exponentially growing technologies are data hungry and until fed well, these won’t yield accurate results.

For more technology insights, follow me @Asamanyakm

Initially, it was big data that helped these evolving technologies. However, you cannot get benefits of big data if fresh, fast-paced, moving data with ongoing purpose is pushed into HDFS, an analytic RDBMS, or even flat files in that matter. You need to understand that processing big data helps you to know what happened at some point in the past, a week ago, a month back, last quarter and so on. It gives a historical perspective, making it possible to say, for example, “This investment in export of an equipment, last annum would have made us X% more productive or profitable, so let’s do it henceforth.” In contrast, fast data is all about the moment and clear inclination towards fast data is becoming visible.

Fast data encompasses extracting insight and creating business value through analytics as well as processing data in motion straightaway as it streams into the enterprise. The source of data can be anything - customers environment, IT systems, financial sources, security infrastructure, IoT sensors, lab equipment, logistics, manufacturing units, field operations, anything you can think of. The gap between resident and fast data represents unused value for any business, and the ability to harness that value will directly impact business success. The secret to success is to act as fast as possible on the incoming data. Fast data is all about that ability to gather, join and transform data in motion, making it fundamental to evolving technologies like machine learning and artificial intelligence.   

Last three days I was at Infocom, the annual mega event by the prestigious ABP Group on business, technology and leadership, attending as a delegate invitee. In fact, the focus of all the discussions, deliberations and debates was on data, AI and ML. For the first time I was hearing Dr. Naresh Trehan, the world-renowned cardiovascular and cardiothoracic surgeon, Chairman of Medanta, speaking on the importance of Data and Artificial Intelligence in healthcare. Since I spent a little less than a decade working on life sciences and healthcare domain from the technology side, I was deeply moved. It is more apt to state that the future of healthcare technology advances resides in medical data itself. Hence the affirmation - data democratization helps Artificial Intelligence and Machine Learning to produce better and more accurate results. In this post, I will restrict my viewpoint to healthcare and life sciences only.

Massive volumes of medical data get generated daily. The more data gets collected, more answers get unravelled, the more doctors and healthcare practitioners learn from it, and the closer they come to discovering revolutionary new treatments as well as cures. However, velocity has always been a challenge. How fast can health data be collected from different procedures, apparatus and observations? How quickly can the data be analyzed to extract new insights? How fast can the concerned practitioners learn as more new data reveals new findings and understanding? How long will it take to decipher and convert all of these into effective treatment procedures or result in drugs discovery that save lives? Until now there was no definitive answer to these queries, but not anymore as technologists working on Artificial Intelligence, Machine Learning as well as Cognitive Computation initiate collaboration with medical researchers and practitioners. Perhaps the healthcare fraternity has realized the problem that the process of correlating and analyzing can take years before they can generate useful solutions from all the gathering data without putting technology to its best use.

Machines have the Capability to learn faster than Humans

It's a fact that all datasets are empirical in nature at their atomic level. Even the most discreet form of data is not open to interpretation at ease. As such, they stay usefully objective and their value increases only when they are added to other forms of empirical data. It’s pertinent to maintain the distinction between this unchanging status quo of data and the subjective nature of human observations, interpretations and opinions.

Effective health care with an emphasis on wellness necessitates the development of understanding based on two very different types of data - structured and unstructured. You get structured data stored in very neat tables comprising of columns and rows. Each column is another field of data, part of a more complete record, each row in the table represents a record. This structure makes it convenient for digital devices to ingest the data and process it. However, as the name suggests, unstructured data poses the enigma that we encounter when handling CT and MRI scans, ECGs and other types of diagnostic imaging. There are images that must be interpreted for the next course of action. The same holds true for written information such as physician notes, diagnostic evaluations, medical research articles, drug formula and much more. Though it was fine for computers to handle records of structured data, it was tough to fathom a computer reading an MRI report, an article or even written notes. Even if a computer could scan and record the image as well as written content, it definitely couldn’t extract insight or draw value from what it was reading or certainly could not evaluate the images, but not anymore.

Three catapulting progresses in technology have induced remarkable new capabilities for digital equipment to collect, analyze, interpret and apply understanding to suggestions that help medical practitioners make better decisions as well as deliver superior healthcare.

Machine Learning empowers computing devices with continuous learning without further programming. The underlying algorithms learn from data and create insight.
Artificial Intelligence works intelligently to be able to make informed decisions based on learning extracted from data. AI resolves problems and makes relevant suggestions.
Cognitive Computation systems learn and reason thereby enabling natural language interaction with people. Using self-learning algorithms that combine data mining, visual recognition, and natural language processing, the cognitive computation systems can solve problems and improve human decision making.

The emerging technologies have empowered computers to read images and written information, evaluate what they have read and scanned, draw conclusions, and make recommendations. No matter how much analysis computers or their human operators perform, and no matter how subjective their interpretations may be, the actual datasets always remain objective that can always be added to and re-analyzed. In other words, datasets can continue to learn, improve and grow.

Application of AI in Real Life

Computer algorithms are making their way into health care at a fast pace from startups and major cloud computing companies. Pager, a startup founded by one of the technical leaders behind Uber, uses Artificial Intelligence to enable its online dispatch system to match a physician and a patient in real time for a personalized home care call. Google, which has recently made a massive commitment to health care through its DeepMind Health initiative, is also using Artificial Intelligence algorithms to enable researchers to better anonymize patient data and find pieces of valuable information to fight disease. In various multi-speciality hospitals like Cleveland Clinic, Stanford, Medanta, and many more AI systems are facilitating remote trainings while surgeons are performing critical surgeries in parallel.

Extracting Real Value from Data

Addition, collection and capturing of data, enables data itself to learn. As they learn they grow in value. New data may often extract new insights that create significant value in improving treatment. Smarter learning systems create more value over time. Most important to remember is how much faster AI, ML and cognitive technologies can accomplish their mission than humans. They process new data into information at rates much quicker than humans. Time to value is reduced, and new discoveries evolve at record speed.

Healthcare Regulation must be Deciphered Correctly

The common misconception is that the Health Information Portability and Accountability Act (HIPAA) represents very restrictive regulations that impede the sharing of information with the objective of protecting the private health information (PHI) of all patients. This is in fact far from reality and further from the truth. Patients are certainly best served when their PHI can be shared readily and securely, where it’s needed, when it’s needed. Intelligent systems now make it possible to properly share specific types of information with first responders who can put it to use. An ambulance crew gets relevant PHI. The fire department gets structural information about the site. And police receive any relevant information regarding potential perpetrators. Most important, each group only gets what it needs, maintaining data privacy

intelligent systems applied intelligently extract, learn, evaluate, grow and enable excellent performance far quicker than humans could achieve alone. All they need to be at best is the availability of right data, at the right time to produce immensely significant and beneficial results.