By Dr. Felix Hovsepian
The term, “Artificial Intelligence” (“AI”) was coined in 1956 by a mathematician John McCarthy, however a more modern definition provided by Dr. Eric Horvitz makes is perhaps better suited to our modern needs. This definition also makes it clear that we’re primarily interested in modeling those aspects of intelligence (thinking) that are amenable to “computation” :
“AI is the scientific study of the computational principles behind thought, and intelligent behaviour”
In addition, Dr. Horvitz describes the 4-pillars of AI:
• Natural Language (communication)
One of the pillars, “Machine Learning” (ML) was first defined by Arthur Samuels in 1959 as,
“A Field of study that gives computers the ability to learn without being explicitly programmed”
and what Samuels had in mind was that we program a machine to learn how to undertake a task from the data (and training) that it has been given. 
There are many kinds of machine learning; one kind is modeled on the neural network possessed by many life-forms on this planet, which include a subset called “Deep Neural Networks” that have been very successful during the last couple of decades. This success has partly been due to the availability of affordable and powerful computing hardware, and massive amounts of data suitable for training such systems.
Nevertheless, these systems are not without their challenges, from both a technical and a societal perspective. Darpa calls these kinds of system, “2nd – wave AI”, and in 2017 John Launchnbury  stated,
“… these systems turn out to be statistically impressive but individually unreliable”
A year later, Darpa announced new funding ($2 Billion) for a new research effort focused on “3rd – wave AI systems”, which would address characteristics associated with commonsense reasoning, perception & human-level communication .
Research in developing intelligent machines have come a long since Turing’s days, nevertheless, we should bear in mind that while current AI systems show much promise, they are but one aspect of a much bigger phenomenon called “The 4th Industrial Revolution”.
Briefly, the 4 industrial revolutions can be described as follows:
• 1st used steam power to mechanize production
• 2nd used electric power to create mass production
• 3rd (or digital revolution) used electronics & IT to automate production
• 4th builds on the 3rd and is characterized by the fusion of technologies that blur the lines between the physical, digital & biological spheres
Prof. Klaus Schwab (the man who coined the phrase “4th Industrial Revolution) once said,
what truly distinguishes the 4th Industrial Revolution from the 3rd is “that it does not change what we are doing, but it changes us” .
We must exercise great care when deciding whether to deploy such systems into human society, because the impact of poorly conceived intelligent machines have the potential to cause individuals (and society) much grief. Notwithstanding, there are also many benefits:
• “ML at the Edge” is the amalgamation of advancements in hardware with ML techniques. Such miniaturized devices can now perform ML tasks at the point where data is collected (detected) rather than having to send the data to the cloud to be processed. Good example includes wearable medical devices, which can monitor various attributes of a host patient and take appropriate action even when the patient may not be able to do so themselves. Dr. Bertalan Mesko describes many examples of emerging technologies .
• Deep Learning systems have recently demonstrated their effectiveness when dealing with x-ray images, in particular when recognizing aberrations in high resolution images – resolutions that are way beyond what an unaided human eye can resolve, and hence why we need the assistance of machines to inspect such these images.
• Modern AI/ML techniques also augment the work of paralegals, and have proved themselves to be useful during the drafting of contracts or when conducting due diligence.
• In manufacturing, ML techniques coupled with smart sensors enable engineers to perform remote & predictive maintenance of expensive machinery.
• In a TED talk, Maurice Conti describes a particularly interesting case study in which a racing car is covered with digital sensors giving the car its own “digital nervous system”. Engineers were then able to capture millions of data values as the car drove around a racing circuit. The collected data was then fed into a “Generative Design” system that assisted the engineers to design a new kind of chassis, which was not only stronger but also much lighter than its predecessors. However, the new design was so complex that it could not be manufactured using conventional techniques, and had to be 3D-printed instead .
• “Conversational AI” systems provide customer care for many businesses, improving customer satisfaction while keeping down costs on 24×7 basis. Such systems also appear in the financial sector, where they are coupled with other AI techniques to provide a service that was previously not thought viable for economic reasons. For instance, Robo-advisors now offer financial advice, automate asset management, and handle insurance claims processing.
• In education, Prof. Ashok Goel has successfully created and used, “AI teaching assistants” during his graduate courses at Georgia Tech .
In conclusion, research efforts involving one or more pillars of AI are underway and will one day result in tremendous benefits for individuals and businesses alike. Nevertheless, we must continue to be cognizant of the hype that we’re all exposed to these days and maintain a balanced view before making a decision regarding the creation and deployment of such systems.
Dr. Felix Hovsepian began his career as a management consultant, returning to academia for graduate studies and research – first in mathematical physics, then in systems engineering and computer science. A career in academia that began with two research fellowships, culminated in a decade spent as a (Full) Professor of Informatics.