Live Q&A transcript: Considering cognitive computing with Judith Hurwitz
Published May 19, 2015
Artificial intelligence has arrived in the commercial world. IBM Watson, robots, natural language processing, and machine learning are all active technologies in dozens of industries. Insights Magazine and Judith Hurwitz, President and CEO of Hurwitz & Associates, hosted an online audience of professionals eager to ask questions about cognitive computing, the most high-profile AI form today.
The following transcript offers the burning questions and considered perspectives that frame the cognitive conversation in today’s economy.
Lee Cook: Can Watson ingest and analyze real-time streamed structured or unstructured data feeds?
Judith Hurwitz: Watson does ingest unstructured data today. For more details I would recommend talking to IBM directly. They are working on a lot of new capabilities related to streaming unstructured data and connecting to structured data sources.
Vaidehi: What exactly is cognitive computing? Is it directly referring to IBM Watson … or is it also for small solutions or small organizations? Are there any other solutions/options than IBM Watson for cognitive computing?
Judith Hurwitz: Watson is not the only cognitive computing environment. There are a few hundred startups in the field that are tackling solutions related to healthcare, financial services, retail and many others.
Edwin: There is an expectation that the user interface and user experience will become increasingly consistent and seamless across multiple devices; for example, for a single session, a user might migrate from one device to another. How do you see cognitive computing being used to support that vision?
Judith Hurwitz: Cognitive computing is a technique for creating solutions that is data-driven. Consistency in user experience is critical to the success of these type of applications.
Guest: In a tele sales environment, how can one develop a roadmap as an online reseller?
Judith Hurwitz: A roadmap is a complicated process. It begins by understanding what you are selling and what your customers need. Most importantly, the roadmap must be able to match the capabilities of data-driven solutions with what is happening in your market. How [to] use data and have that data guide your customers to answers that might not be obvious. The answers are often hidden in the patterns within that data. A cognitive solution relies on collecting knowledge and automating that knowledge through complex algorithms.
Start with the end goals. You want to be able to anticipate changes in customer need and [those] needs to be reflected in your roadmap.
Bruno: How should work organizations mindsets move from structured data approaches to more unstructured orientation? Take financial services, for instance. What valuable use cases have you found that justify a deeper look to cognitive technologies?
Judith Hurwitz: It is true that there needs to be a shift in the way we think about information. In the past, we have never really been able to take advantage of unstructured data in an organized and systematic way. In financial services, there will be ways to leverage unstructured data about a specific type of product offer so that when customers have questions or problems, the system can be trained to understand the meaning of the customer's problem and find potential answers. This is much better than presupposing what customers will ask with FAQs.
In addition, collecting and analyzing what customers are asking helps the company anticipate problems. It can help the company anticipate what new offerings will sell well to customers.
Lee Cook: A case knowledge worker process typically has a case document with metaproperties, a flow of ad-hoc steps—either automated or manual—and supplemental docs data. If I wanted to make…intelligent queries against all the criminal cases against [any name], could I use Watson?
Judith Hurwitz: You could use a cognitive computing system, but there are differences. In a cognitive system, there is a lot more automation that simplifies some of the steps through the use of different algorithms and machine learning techniques. We see the emergence of APIs and tools that help streamline this process. These tools and APIs could include tools that help with metatagging and creating the corpus, as well as doing more automated training
Alan: Should I jump on a cognitive computing project now or wait until the technology matures?
Judith Hurwitz: It is an emerging industry, but I wouldn't wait. I would start with a pilot project because you have to get used to thinking first about data before you just write a traditional program. In addition, there are more and more packaged solutions coming on the market every day.
EJ Lozanta: What is the minimum technology requirement to make cognitive computing and big data analytics possible?
Judith Hurwitz: Because cognitive computing is a cloud-based offering, you are not building a system yourself. It is therefore possible to use some of these tools without a capital expenditure.
Vaidehi: Can you list some examples of cognitive computing solving any business issues/needs different than logically-programmed [systems]?
Judith Hurwitz: The entire cognitive computing model is based on beginning with the data that is available to you or that you can obtain from a third party. This includes everything from best practices to documents with processes and techniques that are foundational to the problems you are trying to solve.
Jeff: How have cognitive systems evolved to assess the patterns hidden in business transactions, social media posts, sensors, and mobile engagements?
Judith Hurwitz: A cognitive computing system is designed based on a foundational corpus of information that is ingested and trained based on pattern matching and a variety of advanced analytic algorithms. The system can ingest a lot of unstructured or semi-structured data. Business transactional data that is structured would have to be handled through other techniques.
Lee Cook: Great answers, Judith! Where can I get a list of Industry specific Watson Cognitive innovations, and can IBM Partners join in the fun?
Judith Hurwitz: I would recommend you get a copy of my book, ["Cogintive Computing and Big Data Analytics"]. It lists many different industries ranging from financial services, drug discovery, security, smarter cities, transportation, retail, and more. The IBM Watson organization has signed up a few hundred partners already in its ecosystem. I was just at the World of Watson meeting in New York last week and there were hundreds of emerging companies doing some really innovative things with Watson.
EJ Lozanta: How close do you think cognitive computing can come to human cognition? What do you think the future holds for the computing model?
Judith Hurwitz: Human cognition is very difficult to replicate, and I don't see us getting there for quite a long time. People like Jeff Hawkins are working on such things, but I think that we can take techniques to help us better understand data and its meaning.
Lucy: What kind of training do I need to undergo to optimize my own contributions in the human-machine interaction that cognitive systems need to perform to their fullest?
Judith Hurwitz: There are two types of training. One is to learn about cognitive computing and the foundational services. The other type of training involves once a corpus has been created and testing whether the data is useful in answering and gaining insights base on the use case. This is an iterative process of testing and training and refining.
Charly: What considerations should I make when evaluating if my company is positioned to engage cognitive computing in a modern, beneficial analytics strategy?
Judith Hurwitz: Think about the data that is part of the lifeblood of your company. Does your company have a lot of written information that is core to what you do? If you had better insights into that data what would the outcome be? Are you really able to understand that data? Can you analyze it today? What are you meaning in terms of predicting customer action or making better decisions? Are you really able to learn the underlying insights from data? That is how you need to think about cognitive computing.
Lee Cook: Does Watson cognitive computing have overlaps with other IBM BI/predictive analytics technologies—IBM Cognos SPSS, Decision Insights? Are there any good papers describing how they fit and interrelate?
Judith Hurwitz: There is definitely an intersection between cognitive computing and advanced analytics tools that you have mentioned. This is the reason that we called the book "Cognitive Computing and Big Data Analytics." It is about the next stage of analytics.
EJ Lozanta: This is a very comprehensive thread, and I am learning a lot. Another question, how would I know if my day-to-day work or projects have areas where cognitive computing can be applied? Also, how can I check if the efforts of applying this computing model is worth the benefits/results I might reap?
Judith Hurwitz: In reality, you will only be able to comprehend the benefits if you experiment. This is a new way of working that is different than many other ways of creating solutions. You are not writing code based on how your business works today. Once you are finished coding a solution, your business will probably be different. Cognitive computing lets your data guide to you to the answers.