Human creativity takes a giant leap in defining new AI use cases for medicine, finance, agriculture, IT, transportation and more.
Humans have always been creative-from harnessing animals and building machines to do physical work, to inventing computers for solving complex problems and offloading mundane tasks. But our creativity has taken a giant leap in the last few years with the commercialization of artificial intelligence, which has applications in medicine, finance, agriculture, IT and transportation, and many other areas.
Artificial Intelligence (AI) encompasses a variety of science and engineering disciplines to make intelligent machines and can be classified into three categories:
- Artificial Narrow Intelligence (ANI) – primarily focused on performing a single task as optimally as possible.
- Artificial General Intelligence (AGI) – capable of deep interpretation of data and the ability to perform intellectual tasks just like humans.
- Artificial Super Intelligence (ASI) – the super-smart machines, which in theory will not only challenge the boundaries of human intelligence but will have the capability of breaking through those boundaries by mimicking social and scientific reasoning.
AI’s core ingredient: machine learning
One of the key techniques that make up AI is machine learning (ML), which is based on the notion of conceptualizing a system model from available historical data that is capable of classifying or predicting new input values that are highly accurate. ML can be classified into one of three forms:
- Supervised learning – Involves the concept of inferring a target function from the labeled input data.
- Unsupervised learning – Does not have a concept of target output or function. Model is generated by exploring the data for any hidden features, patterns or data associations resulting in the efficient clustering of data or recommendation systems.
- Reinforcement learning – Advanced form of machine learning. Drawn along the lines of the reward-and-punishment model in the absence of any historical data.
Unstructured data in the form of text, image, audio or video requires adequate preprocessing steps and must be converted to a structured form before it can be used for any model generation. This is where the concept of deep-learning—which mimics the human brain—comes into play. It involves the process of learning through layers of representation allowing the software to build a hierarchy of concepts out of simpler concepts using models like feedforward, recurrent and convolutional neural networks.
The raw data needs to be meaningfully interpreted for any available information content before ML algorithms can be applied to it. This approach formally extends machine learning into a strategic continuum of analytics, as shown in Figure 1.
Figure 1. Analytics ascendancy curve
The analytics-ascendancy curve triggers with “descriptive analytics”, which entails the application of conventional statistical techniques like measures of central tendency and dispersion to help understand ‘What’s happening?’. It’s followed by “inferential analytics” that works on the concept of deriving conclusions based on induction logic from available sample sets and help answer ‘Why is it happening?’ Feeding more intelligence to understand ‘How’s it happening?’ by discovering patterns or groupings characterized by closely related associations leads to “exploratory analytics”. Ability to make future predictions like ‘What’s likely to happen’ based on learning results in “predictive analytics” and this, in turn, becomes the primary workhorse of machine learning. The most advanced type is termed “cognitive/prescriptive analytics” depicting ‘What do I need to do?’ which eventually takes the form of a recommendation system working on the principle of constraints and optimizations and is leveraged by taking judicious and prompt actions based on the outputs derived from existing analytics types.
Industry Use Cases
One of the focus areas for applying AI has been in the industrial Internet-of-Things (IIoT) domain.
Machine diagnosis and pre-emptive failure predictions
Multiple inventory discovery and monitoring systems are in practice, which gather massive amounts of data from the deployed infrastructure and machines in networks and organizations. More recently, these systems have been used for real-time monitoring and inventory management.
AI offers additional benefits related to preventive and predictive maintenance, asset performance management and the automated tuning of machinery.
The automotive industry is expanding the use of AI, which Gartner forecasts will include 250 million connected cars on the road by 2020. Autonomous vehicles require camera-based vision systems, radar-based detection, and crash avoidance units that use many sensors to keep track of the driving condition, the number of passengers, vehicle mileage, fuel consumption and engine health along with various other vehicle health parameters.
AI is capable of understanding driver features such as emotional analysis, facial and voice recognition and providing virtual assistance with a high degree of personalization and contextualization.
Medical and Fitness Wearables
Medical IoT has created a paradigm shift by providing self-health monitoring and preventive medicine for remote use. Long gone are the days when a patient had to be physically present in the hospital for 24/7 observation under the watchful eye of medical professionals. Now, the data is collected by external sensors or ingested devices that monitor a patient’s condition and ML algorithms that provide the diagnosis with near-instantaneous turnaround time.
In case of emergency, the AI-enabled wearable device can immediately share information with the patient including recommendations of immediate actions and contacts for help.
In addition to these examples, there are many other applications of artificial intelligence that are being applied across many industries and applications, including robotics, recognition systems (for language, speech, and vision), recommendation systems, fraud detection, and software analytics and IT service automation.
The advent of AI has the potential to simplify and enrich our daily lives by transforming everyday devices into complex, adaptive and intelligent systems yielding immediate benefits like a rapid response, enhanced reliability and increased privacy protection. Any misconceptions about the threats posed by these adaptive machines to humans are not only being positively addressed but in-fact culminating into a symbiotic relationship as well.
Accelerating investment in the field of AI is a clear indicator of the forward-looking approach and faith in the potential of machine learning, which harnesses many engineering disciplines. Hence, “machine-learning powered AI” can be categorized as an evolving technology disruptor, and perhaps even signal the dawn of non-human intelligent beings in the making.