The ability to think is one of the things that marks us out as human, “I think therefore I am,” as the philosopher René Descartes put it. But what if machines were able to mimic human thought processes? This is the basic premise behind cognitive computing.
In order to replicate human thought processes, cognitive computing uses techniques like pattern recognition, data mining and natural language processing. The ultimate goal is to be able to make computer systems that are capable of solving complex problems without needing human assistance to prompt them in the right direction. Computing giant IBM which is one of the leaders in the field describes this as, “systems that learn at scale, reason with purpose and interact with humans naturally.”
Research into artificial intelligence dates back more than 60 years to the Dartmouth Artificial Intelligence (AI) Conference of 1956. Over the years this has brought us computers that are able to beat humans at rule-based games like chess, and the first generation of self-driving cars. Cognitive computing, however, holds the promise of taking things a step further, of machines being able to solve less clearly defined problems such as facial and speech recognition, medical diagnoses and sentiment analysis.
How it works
Cognitive computing relies on the use of machine learning. This uses algorithms that are continually fed with data. Over time the system is able to refine the way it processes the data and recognises patters, such that it becomes capable of anticipating problems and coming up with potential solutions.
All of this has a lot in common with artificial intelligence, although cognitive systems have a more complex interplay between their components. Features of cognitive systems include the ability to adapt as information, goals and requirements evolve. They are thus able to cope with unpredictability that would cause problems for conventional systems.
A cognitive system also needs to allow for easy interaction, whether for human users to define their requirements, or with other computers and cloud systems in order to source and exchange data. The system may help to define a problem by coming up with questions or by sourcing additional input where a problem is unclear or incomplete. It may also refer to previous interactions and use information that is relevant to the current application.
To be really effective a cognitive system also needs to be able to understand the context of the data it’s processing. This can include location, time, domain, rules and regulations, and syntax. The system may use multiple data sources, both structured and non-structured, and may also be open to sensor inputs, whether audio, visual or from Internet of Things devices.
A key feature of cognitive systems is that they need to understand the world in which they operate. This means ‘feeding’ them with the knowledge they need to operate. The volume of data needed for this means that we are, currently, a long way from a universal cognitive system. A cognitive application used by an accountant would therefore be very different from one used by a doctor due to the underlying knowledge base required.
Cognitive systems, rather like humans, learn as they go along. They are able to look at the results of an action taken, see if they worked, and apply that knowledge next time a similar problem comes along. Unlike a human, a computer is able to review much more information and its decision making is therefore based on a wider range of evidence.
So, how does this apply to business? One of the first commercial uses of cognitive computing can be seen in IBM’s Watson project. This is a system that was developed to answer questions posed in natural language. Originally designed to answer questions on the US quiz show Jeopardy! In 2011 Watson competed against and beat two former winners of the show.
Whilst winning a quiz show is impressive, more useful applications of Watson are being developed. In particular it’s being developed to act as a clinical decision support system for doctors. It’s able to analyse treatment guidelines, medical record data, notes from nurses and doctors, research materials, clinical studies and medical journal articles.
For businesses there are a number of Watson applications now on offer. These include a Company Analyzer to create profiles of customers, partners and competitors to help with business decision making. Watson also offers Virtual Agents, able to answer customer support questions at any time without the need for human intervention. IBM is making APIs available to allow developers to incorporate Watson’s capabilities into their own apps.
Personal assistance apps are an obvious candidate for cognitive computing. You can already see this in action to an extent with Google Now which is able to engage in two-way conversation with the user and take account of previous questions when formulating its answers.
As the Internet of Things delivers more and more data, companies are likely to turn to cognitive systems to perform analysis and support their decision making. Because of its ability to dig deeper into a wider range of data sources, cognitive computing has the potential to spot things that might otherwise be missed.
Although at the start we talked about the idea of thinking machines, cognitive systems are really about expanding beyond the human thought process. Data is growing faster than humans can cope with and a high percentage of that growth is in unstructured or ‘dark’ data that is almost impossible to analyse using traditional methods. What’s more, according to IBM, more than 90 percent of data will be dark by 2020.
Cognitive systems therefore have the potential to make sense of the huge amount of data generated daily by services like Twitter allowing businesses and governments to spot trends, gain insights into shifts in society and more. Similarly IoT devices like smart meters are generating vast amounts of data which, if properly understood, could revolutionise the energy market and help make the most of renewable generation sources.
It also has the potential to improve security, by looking at usage patterns and spotting anomalous behaviour cognitive security systems would be able to detect and block cyber attacks much sooner than is currently possible.
Whilst it’s still in its early stages, there’s no doubt that cognitive computing will have an impact on all out lives in the coming decades. Added to the greater processing power offered by quantum technology it is likely to completely transform many sectors from manufacturing and finance through to healthcare and transport.