From Machine Learning to Deep Learning

This interview with Elliot Turner, Director of Alchemy and Discovery at IBM Watson, focused on IBM’s cognitive computing technology. IBM Watson’s goal is to use the technology to impact the world, not only by providing businesses with new opportunities, but by improving the world’s healthcare system, helping governments manage risk, enhancing fraud detection, and many other areas. Doing this requires a deep investment, including billions of dollars in capital and hundreds of PhD researchers.

When asked whether Watson was a product or technology, Elliot explained that while celebrities interacting with Watson on Jeopardy might personify it as an entity with almost human characteristics, IBM Watson thinks of it as hundreds of different technologies that they have brought down from the “ivory tower” and given to customers. He explained that it’s also a series of products, a stack of capabilities built upon capabilities.


Elliot explained that there are three critical components required to impact the world of cognitive technology: algorithms, computer, and data. Compute means running massive simulations of how the brain processes and learns. IBM thinks of data, as do their competitors, as the new oil in the world’s economy. If properly mined, data represents massive opportunities.

Many companies have been accumulating large amounts of data assets, but only actualizing a small trickle (less than 15%) in the form of structured data. Unstructured data, such as emails, chats, comments, images, and videos have been accumulated and stored, but have become liabilities because organizations have not been able to actualize and take advantage of the data.

Elliot gave an example of how they are working on helping companies to utilize unstructured data with really interesting outcomes. Weather in a hyperlocal context has a massive impact on the way the world works, such as affecting traffic and certain types of crimes. When a drought in Africa is broken by a lot of rain, it significantly increases the potential for a cholera outbreak. With proper medication, the death rate from cholera is less than 1% death; without medication, the death rate is over 50%. By combining social listening with weather, systems can detect signals from the world about what is going on and identify opportunities for risk reduction, profit, and also impacting the public good.

When asked about private and public data and whether organizations should share data, Elliot responded that although competitive advantage concerns prevent organizations from  coming together, there are safe ways to share data. IBM feels that taking cognitive data and hiding it in a data center is preventing others from reaping its benefits. To this end, they use the Watson Developer Cloud to properly anonymize data and put it on the cloud, making it available to tens of thousands of developers so they can incorporate cognitive into their work. Elliot advised that if you can take part in the shared ecosystem, you should and that we should all work together.

On the topic of how these systems limit human bias, Elliot explained that when they started to learn with systems that were trained by humans, they ran into biases, emotions, and uniquenesses. One of the techniques they’ve leveraged to address this is unsupervised learning. Traditional learning involves a teacher or trainer; however, unsupervised learning is gained by being exposed to the world and deriving a mental model of how the world works. This approach enabled them to scale up reasoning systems and train Watson by exposing it to social media posts. At the same time, they wondered about how human interactions would affect the model.

To research this, IBM Watson created a system that crawled the web, looking at news articles, posts, nearly everything that was written, to form a mental map of the world. After a day, they paused to see what the system had learned. In fact, it had learned hundreds of millions of things, from facts about celebrities to X-ray crystallography. The system also learned that dogs, based on what it had read, were a type of person. Because many people think of dogs as their children, there is certain context in the world that supports the truth of dogs being a type of human. Systems have to be able to have multiple perspectives simultaneously, to understand biases, but also work against them.

Elliot then addressed the major challenges he sees coming in cognitive computing within the next three to five years and how it will be used in new areas. He said that empathy, sarcasm, and the totality of the human experience are challenges that IBM Watson is working on. He predicted that cognitive will be embedded in a vast array of the world’s economy, business, and healthcare, a “dark horse” that will drive a lot of progress and change. For example, cognitive computing will be used to improve medical errors and their impact on public health and mortality, as well as affecting the inadvertent starting of wars. He ended by stressing that while the technology will help people develop better products and services, it is really about making the world a better place.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s