Evolution of Intelligent Systems: Sensors, Smart Data, AI and ML

"A bad system will beat a good person every time." W. Edwards Deming (author of Out of the Crisis, principles for transforming business effectiveness.)

Evolution of Intelligent Systems: Sensors, Smart Data, AI and ML

The convergence of sensors, smart data, artificial intelligence and machine learning is transformative. These technologies are converging to transform industries and accelerate improved outcomes. The availability of highly accurate, affordable sensors is fueling the supply of big data. Analytics and artificial intelligence (“AI”) are converting big data into smart data by distilling an overwhelming amount of data and providing actionable, empirically-based guidance to improve outcomes. Where automation and machines can be applied to the process, smart data for machine learning will continually refine and improve the outcomes in terms of quality, cost, efficiency, time and materials reduction.

~Critical Elements for Smart Systems~

From a business perspective to gain increasing market adoption, the critical criteria for success of the deployed systems include:

1.     Accuracy of the data gathered and reported by sensors and continually improving cost effectiveness of those sensors,

2.     Reliable and dynamic sensors that gather and report data to the system,

3.     Deployment of smart sensors that improve the intelligent sensing of the smart system,

4.     Big data that is intelligent and actionable, effectively “smart data”,

5.     Ingestion and blending of diverse data sources so data (structured and unstructured) can be processed by AI,

6.     Distillation of big data into smart data with visualization and interpretation that saves human processing time,

7.     AI instilled with a Deming-like “constant and forever” commitment to measurement and improvement, and

8.     Machine Learning (ML) driven by AI to deliver results based on data that improve quality, reduce time and waste of materials, report new thresholds of performance for human control of ML boundaries.

~Sensors~

Sensors gather, reveal or measure a physical property and record, report or respond to the information collected. And with improved connectivity and the Internet of Things (IoT), sensors have the capability to dynamically, and in near real-time, report data from remote locations. With Big Data we realize “…extremely large data sets that may be analyzed computationally to reveal patterns, trends, and associations, especially relating to human behavior and interactions.” Big Data Definition

Allied Market Research defines a sensor as “a device that detects, measures or indicates any specific physical quantity such as light, heat, motion, moisture, pressure, or similar entities, by converting them into any other form which is mostly, electrical pulses.”

As the quality of data (measured by accuracy and immediacy,) improve, we overcome the “garbage in, garbage out” aphorism with ever-improving data quality harvested by sensors. The source of high quality, timely input for AI enables performance of tasks normally relegated to humans (or at least human approval) for decisions and actions. We have the computing horsepower and AI solutions to process massive amounts of data and synthesize structured and unstructured data into intelligible insight. That insight creates value by highlighting the opportunities to improve.

Big data, however, is a necessary ingredient but not a singular outcome: making big data “smart and actionable” is the desired result. An overwhelming amount of big data that cannot be efficiently deciphered by humans is of nominal use. Making big data “smart” unlocks the inherent value in accumulating the information. An undue focus on big data is misdirected, it’s really smart data (reliant on big data for input) that has deep value in evolving smart systems.

As we improve the combination of technologies, for example, we improve the capabilities of the sensors as smart machines. Smart sensors harvest information from the physical environment and use internal computing power to perform tasks as they sense certain events have occurred or parameters have been reached. The smart sensor processes this information and relays it to the greater system for action. Proliferation of smart, cost-effective sensors is happening. The 2022 global sensor market is predicted to reach $241B globally, 11.3% CAGR. IoT Smart Sensor

~Smart Data~

Smart Data differs from Big Data as a product of sensors, Big Data and AI/ML that provides. focus and actionable insight. “So, if big data is about volume, speed, and variety then smart data is more focused on creating value, meaning, and accuracy (veracity) for some sort of purpose or outcome.

Smart data is therefore more actionable than big data and thus helps a business function, gain critical insights, or make important decisions.” Academy Smart Data Definition

Big Data + Sensors + AI/ML is an essential input for the creation of Smart Data and with improving accuracy of the data collected and speed of reporting via enhanced network connectivity, the value and predictive ability of the Smart Data is enriched.

~Artificial Intelligence (AI) and Machine Learning (ML)~

AI interpretation of the data simulates intelligent human behavior. AI draws insights from the data that would otherwise depend on human reasoning, labor consuming time, cognitive ability, pattern recognition, experience from similar situations and provides output to recommend a course of action. We enhance the reliability of AI by providing consistently reliable, highly accurate and timely data to help the AI system work intelligently and process more data rapidly.

Couple sensors, data and AI/ML to create a virtuous reinforcing improvement system. AI/ML creates a system that automatically learns and improves from data and feedback without human intervention or repeated human authorization to expand beyond the programmed capabilities. The machines and associated software can access AI-processed data and (within defined boundaries) apply that AI/ML-supplied smart information to learn and improve performance.

AI-enabled devices can process novel experiences by relying on data, analysis, pattern recognition, measure outcomes, measurement of sensed and pre-defined criteria, and self-training within programmed boundaries. By identifying patterns that result in superior outcomes (within the programmed parameters) AI-enabled devices learn without direct human supervision and alter their processes and actions based on data received and processed. Better data in (from ever-improving sensors) facilitates the outcomes of AI-based systems that strive to apply statistical analysis to predict results and measure the effectiveness of the outcomes.

~Convergence of Technologies~

There is explosive growth for the combination and convergence of technologies. Our creativity finds diverse outlets to propagate smart connected devices in new applications and as part of broader systems. The usefulness of these technologies fuels the deployment of solutions that include sensors, connectivity, data, AI/ML. Combinations of the technologies are rapidly under development, deployment and commercialization in consumer electronics, automotive, wearables, mobile devices, healthcare, building management, manufacturing automation and IoT. Sensors in healthcare industry are predicted to be one of the leading growth areas until 2022 with a CAGR of 12.6%. Global Sensor Market Forecast

~Example: Healthcare~

For example, in healthcare AI applies complex algorithms and software to simulate human cognitive process to analysis complex (and often massive amounts of) medical information. With processing power that exceeds human processing capability, healthcare AI systems can rapidly provide healthcare professionals with insights, pattern recognitions based on vast historical databases, and recommend diagnosis and treatment.

When coupled with accurate sensory input and smart data analytics Healthcare AI is a disjunctive change from traditional healthcare technologies as the smart system harvests information that may have been previously untapped and processes it to provide well-defined input for medical professionals. In some cases, by adding machine learning to elements like cushioned hospital beds or surgical tables, an automated machine reaction could take remedial measures to proactively alleviate discomfort or potential injury. With feedback and parameters established in advance through human programming and empirical data for decision-making, the AI input to ML can continually improve the capabilities of the system (smart sensors, smart data analytics, AL and ML). Healthcare AI software is bound by pre-determined goals and, unlike humans, cannot adjust goals independently: the software will perform only within the pre-determined boundaries until such boundaries are expanded by human intervention and authorization.

Healthcare AI software has the potential to recognize trends and behavior to create its own “If this, then that” logic. This, of course, must be subject to rigorous testing and approval before wide spread field deployment to ensure beneficial outcomes. Such a smart system in Healthcare can improve predictive capabilities and empirically-proven outcomes to anticipate problems and recommend (or proactively take) action before harm or discomfort can occur.

Healthcare AI can process data on an aggregated basis and then parse it to very specific demographic applications to alert the opportunity for preventive measures, optimal treatments and patient outcomes. Ideally: with enough quality input, the ultimate Healthcare AI will at a very early stage predict outcomes with enough advance time to take measures to avoid detrimental outcomes altogether. Improved (and improving systems) are already deployed in diagnosis, treatment protocols, pharma development, personally optimized care plans, patient monitoring and care. As the integrated technologies further evolve, we can expect cost reductions, improved patient outcomes, avoidance of injuries (E.g. pressure ulcers and perhaps some diseases like Type 2 diabetes) and ease the burden on healthcare staff who concurrently get the support of accelerated and accurate predictive patient insight.

~Intelligent System Market Outlook~

The converging technologies will compel consolidation and merger in related industries as companies endeavor to break out of siloed solutions and provide a greater, integrated solution for customers. A standalone business in any vertical of sensors, data, analytics, AI/ML will depend on a vastly superior product or services in terms of performance and price, although solutions that bring the pieces together will quickly erode the competitive advantages of integrating disparate but excellent solutions that do not work well with the other essential elements of a complete solution.

Presenting the ease and advantages of a solution that integrates and coordinates all the key elements will emerge as a winning solution. Working in isolation of the other key elements is, for a business, an evolutionary short path.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Thank you to Betty Ledgerwood, Steve Sklepowich and Aloke Gupta for their feedback on this article. I’m grateful that they shared their insights, thoughts and expertise. Any and all errors or omissions are mine alone.

Note: I’m pulling together the discussion points on ethics and the boundaries for AI and ML for a future article. How to prevent runaway innovation by AI and ML without human monitoring has been in the background of many of the points covered. However, it merits greater thought and time than the immediate subject matter of this article: the potential for convergence of sensors, smart data, AI and ML all working as a cohesive, integrated system.