A successful platform has several essential elements, according to Ruchir Puri, chief architect at IBM Watson. Puri, an IBM fellow, spoke with SearchCIO at this year's Platform Strategy Summit hosted by the MIT Initiative on the Digital Economy. In this video, he discusses the essential components of the IBM Watson platform, how he sees the platform evolving and what factors are contributing to its growth.
Below are excerpts from the interview; click on the player to hear the interview in its entirety.
How do you define platform and how do you see platforms changing in the next five years?
Ruchir Puri: For us, the Watson platform is a combination of compute, data and functionality. For platforms in general, the functionality side could be AI [artificial intelligence] functionality, or it could be some other functionality as well. But the various components have to come together in a well-orchestrated way at a global scale and have a level of resiliency and cost-efficiency, which, together, will give rise to a platform. A platform has to be available at a scale and should be consumable by anyone -- not just enterprises, but by anyone in general.
Ruchir Purichief architect, IBM Watson
The first major component is compute -- both infrastructure as a service and platform as a service. Second, is the data side of it and the third is what we will call the functionality side. For the Watson platform, it is the AI functionality. Those three are the major components for how we define the Watson platform.
The way we see a platform evolving is, first of all, the functionality will become much more consumable as we move forward. From the efficiency perspective, as the cloud and the platform evolve, it will become more heterogeneous and specialized. We will start seeing elements appearing in the cloud that are really meant for AI functionality. For example, we've got GPU elements in the cloud which are really infrastructure elements that help AI functionality. We also see the rise of deep learning, which is not because neural networks are new. Neural networks have been around for very long time. However, what is new is the very large amount of data and very large amount of compute in terms of GPUs that have come together to give rise to machine learning models that we couldn't get before.
As we move forward, we will start seeing more specialized compute elements appearing, software as a service built to really utilize those infrastructure elements, and AI services meant not just for big data, but small data as well.