Sign up for our newsletter and get the latest big data news and analysis.

Interview: Fernando Lucini, Chief Technology Officer, Big Data HPE

Fernando LuciniI recently caught up with Fernando Lucini, Chief Technology Officer, Big Data Hewlett-Packard Enterprise, to discuss the future of machine learning at HPE Big Data. Fernando leads HPE IDOL software portfolio and is the global business leader for the Haven OnDemand platform, as part of the Big Data Platform business group in HPE Software. HPE IDOL is the industry’s leading augmented intelligence software for human information and the Haven OnDemand platform powers a new generation of applications with machine learning APIs and Services. He has a deep technical background, and has been critical in the launch of many products throughout his career, spanning enterprise search, rich media analytics, and compliance, among many others. He holds a BEng Hons in Communications and Electronic Engineering from the University of Kent, and an MBA from IE Business School, Madrid.

Daniel D. Gutierrez – Managing Editor, insideBIGDATA

insideBIGDATA: What is HPE’s vision for machine learning? As a follow up, can you shed some light on augmented intelligence and the intersection of the use of technology and humans?

Fernando Lucini: Our vision for machine learning through Haven OnDemand is that anyone and everyone should be able to reap the full benefits of its capabilities. We don’t want to build machine learning offerings that only appeal to data scientists – we want developers of all skill level and expertise to be able to take advantage. As an established enterprise in data analytics, HPE recognizes the true value that data can bring to a business, which is why we have invested heavily in advanced analytics and machine learning.

HPE firmly believes that the future of the enterprise depends on our ability to blend human ingenuity with machine learning. What if people could apply our ability for cognition to all of the data that surrounds us? Of course, that’s far outside of the reach of human capabilities. Thus, what if a machine could do a percentage of the cognition for people, building applications for businesses, for example? We would then be amplifying what experts can do and augmenting that with an intelligent machine, a practice referred to as Augmented Intelligence.

insideBIGDATA: There are many rumors in the marketplace about what machine learning can actually accomplish. Can you help our readers understand what’s fact and what’s fiction?

Fernando Lucini: There is a distinct lack of understanding about machine learning in general. The most common is the fact that machine learning has two distinct pieces; data being the most important one. You cannot fake data. Data must be representative of what’s required, so for any machine learning to get close to achieving its task, whatever that might be, we must have representative data, much like we have a need as human beings to learn. Except machines don’t learn the way we do. The technology we have today makes it possible for people to train machines to learn. The training makes decision-making more flexible and useful than ever before, but it’s still training.

The second piece of the puzzle is the technology itself. In most cases, confusion about the technology is between frameworks/tools vs products that apply the frameworks and tools. If your organization has the scientific and developmental skills in machine learning in addition to ample time (more than people realize) you might use frameworks and tool, as well as hopefully developing IP, to automate processes or otherwise solve business challenges. Alternatively, you can find vendors that provide ready-made solutions that apply machine learning to business problems.

For example, if you need to understand what your customers are telling you in your call center, the first thing you need to do is process the speech. Sure, you can take open source neural network frameworks, 30 scientists (speech experts, mathematicians, researchers) and about 300K of training material that is aligned with your use case and maybe after 3 years you will have your first neural net that can help get text out of speech.

The alternative to this arduous process is to purchase existing tools that apply these techniques and bring unique IP to the mix to deliver superior results with no investment beyond purchase. For this case in particular, a speech to text product would needed.

insideBIGDATA: From what HPE is hearing from customers, what are the biggest challenges that organizations face when implementing machine learning into everyday practices?

Fernando Lucini: The biggest challenge organizations face is cutting through the hype and understanding what problems they can really attack with this technology and make a difference. We are seeing more and more successful use cases with machine learning, which leads us to believe the industry will soon find its water level. Soon we should be able to see an increase in novel uses of this technology.

insideBIGDATA: For organizations looking to adopt machine learning solutions, what should they be doing now to set themselves up for long-term success?

Fernando Lucini: One of the most important factors for success in machine learning is understanding the size and potential of your data. The data you hold today as well as the data you must collect from the present will together make a difference for future insights. While it may be obvious that a company’s data is its biggest asset (together with its clients and products), it might be less obvious that it’s this data that allows machine learning to bring value.

insideBIGDATA: What next for machine learning? Where do you see the industry going 5 years from now?

Fernando Lucini: We are living exciting times and we can see more use cases developing for machine learning. Which, in turn, moves the industry towards greater innovation. In the next five years we hope to see technology that is able to use data with less preparation and thus use substantial flexibility and speed in applying machine learning to all industries.

 

Sign up for the free insideBIGDATA newsletter.

Leave a Comment

*

Resource Links: