IBM has just announced a one billion dollar investment to form a new Watson group. The cognitive computer system famously won the quiz show "Jeopardy!" in 2011 and is now assisting doctors in cancer treatments, works as a shopping adviser and helps with financial planning. In this interview, Manoj Saxena, General Manager of IBM Watson Solutions, walks us through the technology and its possible future applications.
Tell us, how Watson operates and how it has evolved since Jeopardy!
Watson started off as a question-answering machine, but has now evolved into a conversational system. Watson represents a new class of computers that understand and learn from data, including unstructured data, like Facebook updates, Twitter streams or doctor notes. Computers so far have been very good at processing numbers, but they don't understand language and they don't get smarter in time. Watson represents the third generation of computers, we call them cognitive systems that a) understand text and unstructured data and b) learn with every interaction.
Can you give us some examples how Watson is applied today?
Watson is being applied today accross multiple industries including healthcare, retail and banking. In healthcare, for instance, it is being used by oncologists to improve diagnosis of patients and to optimize the time for research to discover new drugs. It works as an advisor and an assistant to doctors. Watson can process all the patients symptoms and the clinical trial data. Today it takes three years for a medical research facility to come up with a new therapy for cancer. MD Anderson, one of the world's leading cancer hospitals we have worked with, found that they can come up with new treatments within 10 months with Watson.
How does the way Watson finds an answer differ from the way humans do?
The biggest difference is that the human mind is unable to comprehend and process all the millions of pages of literature of past test cases. The human memory and processing of data is not geared to have all the knowledge and apply it to a point of decision. Watson basically works as an augmented cognition. It extends the capacity of your brain by giving you more relevant information at the point of making decisions. So it's almost like a GPS system for doctors: If you are driving a car, it gives you alternative ways to get to the destination. And then the doctor chooses which round they want to take.
How does it process the data?
Watson has a technology called the pipeline. Each pipeline is a different learning and understanding model. In the Jeopardy system, we had over 80 different pipelines in Watson. Each pipeline has a different kind of artificial intelligence model to understand the question and then collect the right answers. Think of it as a very highly parallelizable architecture where it takes one question and simultaneously generates multiple alternative answers and options for it that are collected and presented as choices for the user.
What does "understanding" mean in this context?
Watson is understanding, for instance, the language of medicine, it is not understanding the practice of medicine. Humans bring a lot more to a diagnosis than just a machine can, for example things like intuition or judgement. These are capabilities that are not at all within the rounds of the technology right now because we ourselves don't know how the human brain works. So what Watson is great at is semantic understanding of language, being able to pull together supporting evidence and sources and being able to present choices, but it is not good at putting intuition, judgement and other things that are more evolved human cognitive capacities to work. The goal is to augment the human capabilities, not to replace them. It's men and machine, not men vs. machine.
How has big data influenced the development of AI?
I think big data, cloud and mobile have been three significant technological developments that have accelerated the adoption of Watson as an artificial intelligent system. Big data gives the system more data to analyze and learn from, from Twitter Feeds to LinkedIn updates to product reviews. The cloud gives you near infinite capacity to process that data. And then you got mobile which gives you a global distribution channel for people to access that information wherever they are.
What are your plans in developing Watson further?
We are already working on the next generation of Watson. One piece is that we need to apply Watson into more areas where there is a lot of data and content for Watson to learn from. Secondly, we are focussed on adding capabilities to Watson to understand images and videos and not just text. And thirdly, we are now building an ecosystem asking other people to build and invent new and innovative applications on top of Watson. The ecosystem program is part of the one billion dollar IBM investment that will be put into the creation of a new Watson group to scale and commercialize Watson. I am very excited about this ecosystem because – you know as big as we are – we don't have all the intellect and smartness just within IBM, so we want to open it up to the world to innovate on it.
Since DLD celebrates its 10th anniversary this year, which trends and disruptions do you see coming within the next decade?
There are some exciting things to watch for. First of all, the rise of wearable computing. I think this will become an even bigger market than smart phones. I believe cognitive computing is the next big thing that's coming whether you take technologies like Watson, Siri or others and apply it to different areas. Moreover, I think we are going to see some chronic diseases like cancer, cardiac failures and diabetes being attacked by big data and technology and get conquered or at least get addressed in a larger way. And I also think the cloud is going to be incredibly disruptive and a market shift in technology in the next ten years, we will see a whole new cloud for systems. And the last big thing I believe is the internet of things. In this context, security will become even more important and a major battle ground for innovation and risk across countries.
Manoj Saxena will speak at DLD14 conference, taking place in Munich, January 19-21, 2014. Tune in on the beat of our community on the DLD Pulse and find regular updates on the DLD14 programme and speakers here.