Thinking machines

Read time 8min 50sec
Frank Rizzo, KPMG
Frank Rizzo, KPMG

According to Gartner, by 2023, a third of all the work done by specialist lawyers, doctors, traders and professors will be done by smart machines or by less skilled, non-specialist people working with assistance from cognitive computing technology. Gartner also predicts that 90 percent of the jobs we are familiar with today will be replaced by smart machines come 2030.

"Computer systems can automatically detect and interpret what is happening on video surveillance cameras; Siri allows anyone to have a personal assistant in their pocket; Watson has beaten two former champions on Jeopardy and Google driverless cars have driven over 500 000km accident-free. Modern technology is increasingly intelligent," says Suren Govender, Accenture Analytics MD. With the growing availability of sensors, better algorithms for data analytics and growing computational power, these intelligent technologies are becoming more prevalent and are being incorporated in everyday life and business.

"The cognitive era is about thinking itself - how we gather information, access it and make decisions," notes Hamilton Ratshefola, country GM at IBM South Africa. Cognitive analytics engines have the ability to build knowledge and learn, they understand natural language, reason and interact more naturally with human beings than traditional programmable systems. "These systems provide expert assistance by developing deep domain insights and bringing this information to people in a timely, natural and usable way. Here, cognitive systems play the role of an assistant - albeit one that is tireless, can consume vast amounts of structured and unstructured information, can reconcile ambiguous and even self-contradictory data and learn." Ratshefola goes on to describe the partnership between human and machine as one where the two parties are more effective than either one alone.

I Robot?

When we look at the possibilities that can be realised by these innovations, people often express concern that machines or robots running on artificial intelligence (AI) will one day take over the world and rule over humans, says Jaco Rossouw, Principa CEO. "But I think enough of us share this fear that we will always ensure that mechanisms are in place to prevent a machine from ever making a decision that could put someone's life in danger." However, he does acknowledge how the advent of technologies like self-driving cars highlights conundrums that eventually need to be solved. "Do the manufacturers programme a car to hit a pedestrian crossing the street if it means swerving to avoid the pedestrian could lead to severe injury or death of the passenger? These are the kinds of scenarios - the kinds where there isn't a clear right or wrong answer - that must be thought of and addressed when programming artificial intelligence."

Cognitive computing will augment the work we do as humans and not replace it, suggests Frank Rizzo, data and analytics leader at KPMG. This is no different to the automation of factory processes that occurred during the industrial revolution, he says, acknowledging that when we have discussions around automation and digitalisation of processes, humans have already largely been replaced. Cognitive computing and AI become more interesting when we move into more knowledge-based `thinking' disciplines.

"An example would be a medical practitioner who uses cognitive computing to make a better diagnosis - it would be used in this instance to augment a service offering." Some have expressed concern around the idea of machines being able to think and learn, but Rizzo believes there's a huge upside to machine learning and predictive analytics, especially as the world generates more and more data. This is especially true due to the dawn of the Internet of Things (IoT), he adds. "The benefits would be around more accurate and focused results from the analytics due to the machines having richer and more granular data to work with."

Breaking down big data?

For Govender, the recent explosion of data has made it necessary to come up with new solutions to handle these large volumes of information without impacting the responsiveness of systems, but also to take advantage of these information assets to make more informed decisions. The fact that these systems can process unstructured data in the form of millions of documents, then use the information to provide contextually relevant answers, offers unprecedented opportunities across almost all industries - diagnosis (health), product, sales and tariff information (retail), textbooks, papers and academic material (education), he continues.

Do the manufacturers programme a car to hit a pedestrian crossing the street if it means swerving to avoid the pedestrian could lead to severe injury or death of the passenger?

Jaco Rossouw, Principa

With ever-increasing volumes of data, there is a clear need for systems that help exploit information more effectively than humans could on their own, says Ratshefola. Cognitive analytics enable an organisation to rapidly sift through enormous amounts of information, analyse it and supply evidence-based responses to questions. These systems also provide more precise and comprehensive information to customers, which will improve the consistency of responses and reduces response time. Describing cognitive computing as a synonym for `big data analytics' is a mistake, notes Ratshefola. "Cognitive computing is about the ability of automated systems to handle the conscious, critical, logical, attentive, reasoning mode of thought. You can do that, after a fashion, at any scale, in terms of data volumes, velocities and varieties. But you can certainly do more powerful cognitive applications-such as natural language processing, sentiment analysis and streaming video object recognition-at scale."

One of the misconceptions around cognitive computing is that it can only be applied to unstructured data, states Govender. But this is not the case. Cognitive systems are equipped to harness valuable insights from unstructured data and leverage off structured and semi-structured data. These innovations can be used in any situation where the implementation makes sound business sense.

While he is adamant that cognitive computing is a game-changer for businesses across every industry, Govender cautions that organisations shouldn't become too enamoured with any particular technology, as if that technology by itself is the answer. "It is vital to think first in terms of types of work, and then consider the business rationale for integrating technologies into a total cognitive computing solution related to that work."

Predicting points

Sports fans make predictions on match results based on their experiences, observations, gut feel and personal bias. Machine learning is a little less emotional, notes Principa CEO Jaco Rossouw. As a side project, the Principa team recently used predictive analytics and machine learning to predict the results at the 2015 Rugby World Cup. According to Rossouw, the predictions were based on data extracted from 6 000 matches played by 99 teams since 1995. As the tournament progressed, the results from each game were added to the model so it could adjust the forecasts according to the recent tournament developments. At no point was the model concerned with what teams were actually playing the games, he notes, but rather the characteristics of each team. "So there was no bias towards any team - it was just looking at the data or the characteristics in making its prediction."

Cognitive computing is about the ability of automated systems to handle the conscious, critical, logical, attentive, reasoning mode of thought.

Hamilton Ratshefola, IBM South Africa

Whenever they added new data, the model would adjust expectations and the accuracy of the predictions increased with every match played. "For machine learning to be effective and accurate, it must have a great amount of data. Decisions based on machine learning get better or more accurate with time, as they learn what decisions and actions lead to a negative or positive outcome." One interesting set of data, which added a bit of the human element to the predictions, was the inclusion of fantasy player values and bookie odds, notes Rossouw. "We didn't look at fields, or influencers, such as the referee or the weather, as this would not necessarily have resulted in a large enough increase in accuracy of the prediction."

Cognitive and the customer

Amazon's recommendation engine uses big data gathered from its database of around 250 million customers to suggest products by looking at a customer's purchase history, what similar people have purchased and a multitude of other variables. This allows them to understand their market and use the data at their disposal to give their customers what they want. In the past, sophisticated marketers would work through the demographic information of their customers and then assess the psychographics of these people to create personas that tie into a specific brand's target audience. Cognitive analytics makes this simpler by allowing marketers to dig even deeper, uncovering meaningful consumer insights and shared characteristics so they can map consumers into personalised groups.

"By being able to collate and evaluate data around this marketing aspect, it will assist all marketers to build better brands and connect more with their customers," says Stuart Innell, senior strategist at Boomtown. "Cognitive analytics can and will fundamentally change the way humans and marketing interactions take place. The information will lead to developing deep insights around aspects that most marketers have only touched on."

Consumers are complex with various habits, preferences and characteristics, he continues, noting that the potential for the clarity and accuracy of information provided by analytics, which could be fed into building campaigns and brands, is highly valuable. And these innovations span industries; healthcare organisations are using predictive modelling to assist in the diagnosis of patients and identifying risks associated with care, and farmers are using predictive modelling to manage and protect their crops from planting through to the harvest. "Big data technology allows companies to gain the edge over their business competitors and, in many ways, to increase customer benefits - but only if the ability to analyse and interpret this data is available. So the partnership of machine and human ultimately needs to exist."

This article was first published in the April 2016 edition of ITWeb Brainstorm magazine. To read more, go to the Brainstorm website.

See also