If 2015 was the year of Cloud, Analytics, Mobile, Social and Security, 2016 was the year of Cognitive.
Cognitive is the big word at IBM right now so I thought it would be an appropriate time to share my understanding of cognitive computing and how I see it relates to business and technology.
Firstly I hear you asking, what does cognitive even mean?
Well, cognitive is the word which is used to describe systems which can understand, reason and learn. Like all IT systems, cognitive systems rely on data, which is exponentially increasing every year – in fact nowadays most businesses probably have too much data – and more importantly aren’t sure how to best go about creating value from it. Cognitive systems help to reveal patterns and relationships across data which could previously not be discovered by other means.
As the explosion of data gets ever bigger, consultancies such as IBM are helping their clients understand and refine their data better by building or integrating cognitive systems into their existing systems so they get better value from their data. And it’s not just IBM – Microsoft has Cortana, Facebook is developing chatbots, Amazon has its Echo product which comes with its own assistant Alexa, and in fact you could argue those of us with iPhones already have pocket cognitive assistants in the form of Siri, so Apple are in on it too. It’s safe to say cognitive is a competitive market already.
Those IBMers reading may well (and should!) know that Watson is IBM’s lead brand in cognitive. Watson (the question and answer based computer system) famously beat human contestants on the US game show Jeopardy in 2011 which gave way to its own dedicated business units shortly after. Today, thousands of people work on Watson and its various umbrella products such as Watson Analytics and Watson Developer Cloud at IBM across the world.
So how does it work?
At a high level, Watson is a mix of different systems and technologies – leveraging artificial intelligence, machine learning and natural language capabilities. It understands questions asked from end-users – in a variety of different languages! – by using semantic parsing, which is fancy for breaking up the words in the question to understand what the question is asking. It then works off a “corpus” of knowledge, which is made up of structured data, such as tables and databases, and unstructured data, such as articles and audio. Watson then uses advanced computational techniques to very quickly extract information related to the question from its corpus.
Finally, it analyses what it understands it has been asked against what it knows to be correct and produces answers with a confidence level of accuracy – a bit like a super-powered search engine – but you don’t have to go through each result to find what you are looking for, Watson can do this for you which is what makes it so powerful.
However for this to work, firstly a given instance of Watson to be used must be “trained” using a variety of sources – human experts teaching it the right answers to questions, machine learning techniques, and feeding interactions between real life users and Watson into its corpus of knowledge. It also needs to hold data based on the questions it may receive. The advantage of this is that Watson can become very specialised in a certain areas, for example medicine or law. The disadvantage is that it requires a lot of time (and money!) to train it to answer questions and know what is actually correct because well, even Watson can’t know everything.
So overall, Watson is a system which runs analytics against a body of data to get insights from what it has been taught to be correct from experts, evidence, and real interaction, and can estimate a level of confidence to potential responses to its requests. The end goal of this technology is to inform experts to make better decisions, not to replace them, especially in data intensive industries.
The reason why these new-age systems are called cognitive is that they display very similar capabilities to the way the human brain responds to requests to do something. For example every time you must make a decision to do something you are making it based on from what you learnt from experts (teachers, education), from evidence (what you know to be correct, your values), and from interactions (your life experiences) and respond accordingly.
Already IBM has exploited Watson in almost every industry, notably in healthcare and automotive industries. At a recent conference in Las Vegas called “World of Watson”, IBM CEO Ginni Rometty and a Japanese professor announced that by using Watson they discovered a patient was suffering from a different, rare form of leukemia by feeding the patient’s data into Watson. Ultimately they were able to treat the patient accordingly after doctors by themselves were unable to come to a diagnosis, therefore potentially saving the patient’s life. How cool is that!
It’s all very well having this cool technology, however to gain approval for projects involving cognitive, there must be a good business case. In my opinion, these are the main benefits to business:
Customers have better engagement experiences with cognitive systems. To illustrate, I’ll use an example of a situation that probably frustrates all of us. When dialling customer service to get a quote from an insurance company, instead of being put on hold for half an hour and losing a potential customer, we could access a cognitive chatbot who could handle the initial interaction and gather the simple details that every quote needs, speeding up the time taken to handle the quote. But if it got into any trouble or could not handle the interaction we could be put through to a human user.
Decision makers are better informed by cognitive systems. Business leaders can gain relevant insights, tailored to their industry, from cognitive systems helping them understand more about their customers, markets, and what people are saying about their products and services from different sources of data – and respond to this information, often in real time.
They help us to explore and discover about the world and ourselves even more. Cognitive systems are making their way into the healthcare industry, assisting doctors and nurses to keep us healthier, and travel companies are creating tailored experiences for each customer by pairing cognitive systems with the vast amount of data they hold about different countries, cities and travel options like flights and trains.
It’s obvious that these systems do have potential, however we must be careful in the way we use them. What they can’t do is analyse the risk that might not be represented in the data such as environments, people and cultures.
There are still limitations to these technologies – e.g. if a predictive model suggests to buy oil in the Middle East because it is likely to be cheaper than buying oil elsewhere in the world, but the country is in conflict and its leaders are in danger and is not represented in the model, then this must be factored into the decision – in the end you can’t hold machines accountable for bad decisions.
Furthermore, I would say that smart as computers may become, there will always be a humanity gap between humans and computers, as they will always lack fundamental human abilities such as consciousness and those intangibles things which make us human such as feelings and thoughts.
So there’s a way to go before we see the film “I, Robot” come to life.
Given the rapid pace of innovation in the tech industry, I also believe that cognitive systems will become more accessible to the public not too far in the future too, just in the same way that the smartphone did in the previous decade.
So overall, I see exciting times ahead in technology at IBM and at the cutting edge in this industry. Happy December to everyone and a very Merry Christmas to you all! See you in the new year,
Richard Cure.