IBM and the milestones in artificial intelligence 


(Published in The Produktkulturmagazin issue 2 2016)

Cognitive systems can change the way companies will be thinking, acting and working in the future. With Watson, IBM is developing a semantic search engine that can collate and answer questions asked in natural language. In the future, this kind of software could support us in many areas, such as in making complex decisions that have to be made under extreme time pressure. Dirk Heitmann, Director of Cognitive Solutions for Germany, Austria and Switzerland at IBM, tells us just how this works and how Watson could otherwise be deployed in the future. 

Mr Heitmann, what significance does Watson have for the further development of cognitive intelligence? 

Dirk Heitmann: Watson is a pioneer. This cognitive system was first unveiled to the public in 2011 within the context of Jeopardy!. The event marked the beginning of a new era of cognitive computing. In this US-American quiz show, quick thinkers with extremely broad knowledge and the highest level of language competence compete against each other. They do not have to answer any questions, but formulate the correct questions to complex answers – a considerably more difficult undertaking. For example: “He is the father of bacteriology.” The correct answer to the question is: “Who is Louis Pasteur?” With this quiz, the answers are often full of double-entendres, metaphors, irony, foreign-language terms and synonyms – and these are precisely what generally confound computers. Even today, and also if they understand language. Not Watson: it beat the two all-time Jeopardy! champions, ultimately by a huge margin. With this, the system proved that it has acquired skills that no other system has been able to reveal to date. And it has demonstrated the potential inherent in such learning systems. 

What challenges could be solved by Watson in the future? Where do you see further development potential? 

D. H.: Watson has been conceived to process and evaluate data, regardless of their origin or the form in which they are presented – also pictures, videos and voice files. This is new. Because to date this information, which actually makes up around 80 percent of all available data worldwide, has not be usable for computers. As a result, cognitive systems such as Watson offer important practical assistance in saving lives, treating disease, developing better products and increasing the efficiency of companies. At any rate, we are convinced that this technology represents our best – and possibly our only – opportunity to solve some of the greatest problems on the planet: from successfully treating cancer, climate change, all the way through to better understanding complex business correlations within the context of the Internet of Things (IoT). In general, we are only at the very beginning of this development. 

How do you respond to people with a negative view of cognitive intelligence? 

D. H.: Cognitive systems are being developed to help people and provide them with support in their day-to-day work. However, the final authority should always be a person. But there is a further aspect: they can close gaps where services are not, or no longer, offered as a result of a lack of profitability or availability. To this end, the Watson financial consultants would be able to competently advise bank customers with small capital assets, hence taking on the task that is fiscally no longer feasible for the classical investment adviser at a bank. 

Watson may be a machine, but it does have some human characteristics, such as the ability to understand irony and cynicism. How exactly does this work? 

D. H.: Watson has no human characteristics. However, it is – by means of intensive training – indeed capable of learning
certain things, which actually happens to include identifying irony and cynicism. The basis for this is a new generation of algorithms and man-machine interfaces that permit the system to process structured and unstructured data in equal measure, to identify patterns, generate correlations and hidden interactions, and hence also to develop its own understanding of topics and issues. Here, the system works with neural networks, traditional machine learning, text analysis tools, voice recognition and currently around 50 different application-programming interfaces (APIs). These are interfaces with which Watson is provided and trained with special knowledge, on healthcare, financial subjects or technical expertise, for example. 

If the information basis is appropriate, Watson is also able to compile forecasts of the most diverse kind. What precisely can it forecast and just how complex or accurate can such forecasts be? 

D. H.: The more intensively and comprehensively Watson is coached, the better its results and forecasts. The system was, for example, coached to identify and correctly attribute written customer complaints on behalf of a large German insurance company. To this end, Watson had to analyse and correctly assess thousands of texts with the help of its coaches – insurance company employees and linguists. In the process, the learning system was conditioned to attribute the sentences in the texts to predefined categories. It identifies these now with almost one hundred percent accuracy and automatically sends the letters to the correct processing clerks. Another very current example comes from the area of IT security. Here, we will – together with 200 IT students at well-known US-American universities – be coaching Watson in the autumn of this year to identify cyber attacks at a very early stage and provide recommendations for countering them. For this, the system is fed with knowledge and findings relating to IT security, with up to 15,000 documents each month in the initial phase – including information on malware or even entire databases on the history and successful tackling of cyber attacks. This training is conceivable for various disciplines with the most diverse objectives – also for accurate forecasting of weather developments, for instance. For this, we brought ‘The Weather Company’ on board some time ago. 

What are the next technological milestones for artificial intelligence? 

D. H.: What we are currently seeing and experiencing is merely the tip of the iceberg. We are still at ground zero. Because what is happening right now is pushing the boundaries of available knowledge into the virtually endless. With this, the potential of such technologies is equally endless, and trying to define and determine the next technological milestones is virtually impossible as a result of the exponential development of knowledge and its abilities. We, at any rate, have the world’s largest commercial research department and continually collaborate with universities and partners in the most diverse industries on expanding Watson’s abilities. 

What areas of application in which sectors of industry will be in the spotlight here over the coming years? 

D. H.: IBM is working on the development of cognitive solutions with a huge number of partners from the most diverse sectors. We have realised initial commercial applications in collaboration with the US health authorities. With good reason: because the ‘problem’ with medical progress is its incredible speed. Not even highly-specialised medical consultants are today able to keep up with all scientific publications in their area of specialisation and remain on top of progress and developments. And general practitioners, frequently the first port of call for diagnoses, do not even have a remote chance of success here. In this regard, a cognitive system can support doctors in making the right diagnosis faster. But the journey continues, as the already mentioned examples showcase: in the manufacturing sector and in the automobile industry, the health care sector, in finance, logistics, trade, pharmaceuticals and in the raw materials industry. 

To this end, can Watson support us in utilising existing resources more carefully? 

D. H.:  This is undoubtedly a very central aspect of the deployment of learning systems such as Watson. For example, Watson can help optimise the feeding of renewable energies into the electrical power grid or regulate heating or air-conditioning systems in buildings more precisely by evaluating the corresponding weather data. And the kerosene consumption in aeroplanes can also be more accurately calculated in advance using Watson. Furthermore, the deployment of raw materials in production processes can be better planned, or value chains and networks can be managed considerably more efficiently. To this end, a learning system can, among other things, provide a timely warning in the event of delivery bottlenecks or – conversely – when stocks need to be reduced. With regards to the utilisation of resources, the principle is in any case optimisation, and that always translates into more careful handling. 

What would be your personal favourite scenario for a future deployment of Watson? 

D. H.: I personally think it would be truly fascinating and exciting to use Watson as a personal assistant and advisor in my everyday life, helping me make well-founded business decisions faster. And once Watson has learned to identify humour, this would undoubtedly be an exciting and enriching cooperation. 

Loved this article? Put it in your cart!


Since March of this year, Dirk Heitmann has been heading up the Cognitive Solutions team for the Germany, Austria and Switzerland region. The department bundles those IBM activities focussing on cognitive analytics solutions, including Watson, for instance. At the same time, he is also responsible for the automobile industry, aerospace and defence within the company’s Cognitive Solutions team.

Picture credits © SoftBank Robotics/Jake Curtis

Leave a comment

Please note, comments must be approved before they are published