The evolution of AI: Why it’s finally time for business to get excited

Group of vintage robots

AI as a practical concept

Colloquially, the term “artificial intelligence” is applied when a machine mimics “cognitive” functions that humans associate with other human minds, such as “learning” and “problem solving”. — Wikipedia

Perhaps useful for academic purposes, this Wikipedia definition of artificial intelligence (AI) doesn’t really reflect how AI will influence us over the next 10 to 15 years. The practical reality is, AI is a range of technologies — a more intrinsic intelligence within systems that leverages data, information and knowledge, and incorporates learning.

Early days

In 1950 Alan Turing laid the modern-day foundations of computing and AI. Turing introduced the concept of the Turing Test, a revolutionary idea in which an AI would be judged intelligent if another human could not tell the difference between the responses of a human and the AI machine.

Early forays into AI were fairly rudimentary, and a lot of effort was put into developing speech capability. This has reached the point where speaking is now commoditised — all modern computers and smartphones can speak text. Speech recognition is also pretty good for limited grammar.

In parallel, expert systems were developed to mimic human intelligence, but these required complex programming languages. These were not particularly intuitive or flexible, and they mostly failed at developing AI. However, over time expert systems became “Rules-Based Systems,” which are widely used today.

Moving to neural networks

By the late 1990s, research on expert systems declined and the focus shifted to neural networks, the basis of machine learning and deep learning. Both are primary tools of AI today, as the ability to learn is key into making AI actually “intelligent.”

In its simplest form, neural networks are computer programs modelled on the neuron structure of the human brain. Artificial neural networks are typically designed as layer upon layer of neurons, where the nodes in each layer are processed in parallel to maximise performance.

The number of layers of neurons is a key differentiator between machine learning and deep learning — 3 or 4 for machine learning versus over 150 for deep learning. Another key difference is that deep learning networks are typically capable of learning by themselves, whereas machine learning networks require humans to train them.

Recent times

The enormous storage and computing capacity offered by the cloud has allowed neural networks to finally be accessible to the public and enterprises, instead of being restricted to a small number of experts. This evolution towards “practical AI” and neural networks is one of DXC Technology’s key trends for 2018.

The reduction in storage costs has facilitated the construction of large data sets for training deep learning neural networks. For example, the Google Open Image dataset contains 9 million images and details of their content.

However, most enterprises are not interested in differentiating between images of a cat and a dog. They would rather that AI do things such as:

  • Find imperfections in manufacturing processes based on the pattern of data automatically collected from the field
  • Identify deformed fruit on a conveyor belt
  • Address customer queries asked in natural language autonomously
  • Detect an insurance fraud during a phone conversation based on the analysis of the voice and vocabulary used by the person lodging a claim
  • Find expert information buried in an unstructured set of documents
  • Anticipate potential damage to a mining conveyor belt before it fails

AI today are capable of achieving these tasks and more. The AI systems created with building blocks allow access to highly complex capabilities easily in order to build what is needed. Examples of building-block capabilities are:

  • Understand natural language to have conversations via text and speech
  • Identify emotions through voice analysis
  • Find answers to common questions as well as those more complex
  • Recognise images
  • Teach (train) the AI for the specific context it needs (e.g., recognising a cat from a dog is different than recognising a small defect on a conveyor belt)

What does AI mean for enterprises?

While it’s only the beginning of the journey, AI systems today are starting to be truly “intelligent” and present huge opportunities for disruptions for enterprising companies. This potential can be exacerbated when used in conjunction with other technologies such as drones or robotic process automation.

Some specific industry use-case scenarios we’ve encountered include:

Manufacturing — Image and video analysis of manufacturing processes is an area of significant growth. Deep learning neural networks are best suited to tasks such as infrared and visible light image processing as part of quality assurance processes. For instance, AI coupled with a video system can be used on a manufacturing line to detect quality issues.

Customer service — Access to an automated customer service virtual agent through text and voice channels allows companies to improve their customers’ response time as there’s no need to wait for an actual person to be available. There is an increase in operations hours as virtual agents work on a 24×7 basis at no extra cost (reducing overall operational costs). Further, by ensuring the consistency of responses, the quality of service also improves.

Insurance — Image analysis of property insurance speeds up processing by recognising straightforward cases and processing these without human intervention. AI can analyse tone and voice pattern of customers lodging claims to help customer service agents to detect fraud cases.

Asset management — Autonomous drones can fly along linear assets such as power lines, railway lines, waterways, water and gas pipelines, or even roads, taking a continuous stream of photos or video. A trained Deep Learning solution would enable assessment of maintenance requirements at a massively reduced cost and, more often than not, at reduced risk to humans — such as inspection for cracks in windmills tens of meters high in the air or pit holes in roads in remote areas.

Health — IBM Watson is famous for helping doctors find the best possible cancer treatment based on each patient’s specific case from the vast amount of documentation written on the topic. In Australia, Watson can detect melanoma with an accuracy of about 95%, better than the 80% that the average dermatologist typically achieves.

Emergency services — Data and images from drones combined with AI offer off a variety of benefits in natural disasters including flood, fire and rescue. Currently, accurate prediction of floodwater peaks is a prerequisite for effective flood management, but there is limited instrumentation of inland waterways to monitor this at present. Using AI, drones could fly GPS-mapped flight paths capturing a stream of images for conversion to a volumetric model, which could be compared against data captured during flood events. The information derived from the baseline model and the flood water volume could then be used to project forward in time to determine the potential impacts on people, livestock, property and the environment downriver.

It is truly time to get excited about AI. This is the beginning of a significant technology shift. In 5 years, AI will be embedded in some form or shape in every single app and system, and in 10 years it will be business-as-usual to do it and nobody will think further about it. Companies who don’t want to be left behind need to embrace the change now.

Happy AI-ing !

This article is the first of a series about AI from Focus the Way Forward. Watch out for our next fortnightly installment!

Paul Hamilton is a Sydney-based public sector consulting lead with DXC Consulting. Holding a Master of Science in Computing from the University of Technology Sydney, he has more than 30 years of experience delivering complex systems integration solutions and business consulting across, public sector, manufacturing, utilities, telecommunications and defence.

Antoine Giraud is leading the IBM Practice within DXC Technology in ANZ. He has a Master of Engineering from the Ecole des Mines in France and has nearly 20 years of experience delivering IT consulting solutions in Asia and ANZ across industries including public sector, manufacturing, utilities, telecommunications, facilities management, oil and gas, banking and defence.


2018: AI gets smarter and more practical

Automation and customer service — so perfect together

A successful digital collaboration transformation starts with your users

How smart cities can engage citizens and bridge the Digital Divide


  1. […] article is the second of a series about AI from Focus the Way Forward. Watch out for our next fortnightly […]

  2. […] The evolution of AI: Why it’s finally time for business to get excited […]

  3. […] The evolution of AI: Why it’s finally time for business to get excited […]

Speak Your Mind


This site uses Akismet to reduce spam. Learn how your comment data is processed.