For decades, results of Artificial Intelligence (AI) research remained relatively unimpressive. But the creation of viable AI products, and this technology becoming more mainstream have changed the way we look at smart machines. We've seen the headlines: Deep Blue, IBM's supercomputer, defeats chess champion Garry Kasparov – 1997; IBM's supercomputer Watson trounces its two competitors on television – 2011; Google's AlphaGo beats Go master Lee Se-dol – 2016. This technology is definitely arriving in a hurry. And if predictions on the rate of adoption over the next few years are accurate, this will transform our lives in many profound ways.
This article will present a definition of Artificial Intelligence or Cognitive Computing (CC) as it is better known today, and some categories of current applications of the technology. It will also suggest how the technology might affect the defence business, either benefiting our communities or threatening them.
What is Cognitive Computing?
There are no agreed definitions of CC; however, we can identify the following commonly acknowledged characteristics:
1. Natural Language Processing – The machine should be able to understand everyday languages, in the spoken or written form. This means that the user does not interact with it using special 'programming languages' or even menus with finite choices. The machine is able to deal with data that does not possess any predefined structure.
2. Machine Learning / Deep Learning – Normally machines are programmed by experts, using programming languages which command the machine to behave in specific ways. A CC machine is rather programmed in a manner similar to a human being, it learns based on its experiences.
3. Perception – The machine needs to be able to take into account its changing environment. It has sensors that extract information which it uses to learn and make decisions. These might include: sight, sound, speed, temperature, network probes, etc.
4. Mimics Human Abilities – The machine behaves in ways similar to a human. It does not have a predefined set of rules which it applies to a problem, but rather it reasons, defines hypotheses, and makes recommendations as to appropriate courses of action. This can lead to unanticipated and novel results.
This categorization of Cognitive Computing is intended to discuss the types of applications of this technology which are being made today and in the near future.
1. Category 1 – Intelligent Personal Assistants. We have all used Cortana, Siri or Google Assistant on our mobile phones to ask questions, play our favourite tunes, or switch on our household lights. These are much more than voice recognition applications as they are connected to other applications and respond to our commands.
2. Category 2 – Specialized Applications. This category consists of specialized, and somewhat limited applications, designed to solve particular problems. One interesting example is x.ai's Amy, an application designed to schedule meetings for you. Once you've indicated the meeting's participants, it searches all their calendars to find a slot when everyone is available.
3. Category 3 – Intelligent Agents. These applications use cognitive technologies to replace people, carrying out repetitive low to medium skill jobs. Examples include ipSoft's Amelia, a cognitive agent that replaces / augments call centres and is able to answer calls, diagnose problems, resolve issues or escalate the call to the next support level. Amelia communicates using everyday languages, works 24/7, senses the emotional state of the caller, and can even help defuse a situation.
4. Category 4 – Platforms. While many companies are working with the technology to build applications to solve a narrow class of problems, others are working to create general-purpose cognitive platforms that can be used to build solutions to a wide range of problems. This includes, among others IBM's Watson. Applications such as GoMoment's Ivy have been built on top of Watson to satisfy a particular need – in this case, a 24-hour hotel concierge service.
5. Category 5 – Robotic Agents. This is an exciting, rapidly developing category that is likely to have a very profound effect on our lives in coming years. It differs from Category 3: Intelligent Agents, as it mixes CC with a physical presence and mobility. Self-driving cars will soon be on our roads, there are hotels in Japan where everything from the concierge to the bell boy are robots, reducing labour costs by around 75%.
Opportunities. It is likely that we will see the introduction of these technologies into NATO systems in the coming five years. Some Nations are already using cognitive technologies for specialized analysis, particularly in the intelligence world or to supplement call centre staff.
One of the first areas we might see adoption is in the Centralized Service Desk (CSD) type of applications. Using technology that is available today, we can provide a 24/7 service that can resolve perhaps 50% of help tickets. This is in line with what is reported by large banks and other institutions that have adopted this technology. This can save money and increase service levels.
Cognitive technologies are very good at analysing large amounts of data and identifying anomalies. In our world, this could include automated extraction of intelligence data from video imagery, pattern analysis of maritime traffic looking for unusual behaviour – perhaps indicating smuggling or piracy – or looking for patterns of behaviour which could help identify insider threats.
Another area where we see this technology being adopted in the short term is in healthcare. These tools can be used to assist in a diagnosis, reducing the possibility that overworked or fatigued doctors might overlook symptoms. A doctor will still be involved, but as a tool assisting the doctor to make better diagnoses, this technology has great potential. We may see this in our medical applications and field hospitals.
Threats. As smart machines become increasingly capable, they will become viable alternatives to human workers under certain circumstances, leading to what has come to be known as 'Virtual Talent'. This is likely to have a profound impact on the labour market. This may lead to a utopic society where robots do all the work and we enjoy much more leisure time. However, it is equally likely to lead to high rates of unemployment and unrest. This may create tensions in societies affecting our business.
"A super intelligent AI will be extremely good at accomplishing its goals, and if those goals aren't aligned with ours, we're in trouble." - Stephen Hawking
We also have seen new types of cyber threats emerge as these technologies have proliferated. Because they learn from their experiences, they can be taught bad behaviours as well as good. There was a recent case where Microsoft released a 'chatbot' via Twitter named Tay that could reply to simple text exchanges in an intelligent way, providing a virtual friend. Within 24 hours of its launch, people had taught Tay behaviours associated with the Nazi regime and so it was decommissioned by Microsoft.
Well-respected scientists and Industry leaders are cautiously warning about the potential negative impacts of these technologies, prophesizing a danger to our very existence. As machines get smarter and more independent, it is possible that they will start to make decisions and take actions that are not to our advantage. Industry leaders such as Elon Musk are suggesting that there should be some sort of regulation of this technology to ensure it benefits mankind.
Conclusions. Cognitive Computing is advancing rapidly and will affect us in many predictable and unpredictable ways. This advance cannot be stopped so we must be ready for it - either to adopt it to our advantage or to be prepared to defend ourselves against it. With the open sourcing of Google's TensorFlow, this technology is within everyone's reach – for good or bad.