We have big problems to deal with as a country, cost overruns, ramping inefficiency. And now we have tools that can help us deal with that, where I think 10 years ago we didn’t; we had relatively primitive tools. So, I’m excited for the next decade. I think that we’ll be much better doctors, we’ll hopefully get much better health care as patients, and some of these tools will help us get there. I think when people imagine machines and people working together, sometimes it’s a little frightening idea of having a computer help you think. But I imagine it’s being kind of like a violin, that if you look at a violinist and violin together, it’s really the violin that’s making the noise. But the violin and the violinist are able to do something much more than either of them could do separately. I think the future is going to be humans and machines working together like that, like the violinist and the violin. – video ends]
A nice little thought there at the end of how these systems are going to continue to evolve to become collaborators, as we talked about earlier. So, as we think about the future, IBM Research will continue its groundbreaking work in the area of cognitive computing. We’ll continue to get the next wave of things ready for my teams. And Guru Banavar is going to come help us understand what the other side of the looking glass looks like. Thank you. Guru?
Guru Banavar – Global CTO, IBM
So, as a computer science researcher, a career computer science researcher, this is actually an inspiring day for me, because many of the things we’ve been working on for decades is now mainstream, as you say it, Mike. So, I’d like to give you a sense of all of those, actually step back a little bit and talk about what’s been going on in this field for quite some time and what is likely to happen.
Now, we’re talking about decades scales here. So, Watson winning Jeopardy may have seemed like a sort of, you know, it came out of nowhere, but really there’s so much of foundational science underneath it, computer science underneath it. There’s a lot of work that’s gone on in academia, but I’m proud to say that IBM Research has actually invested for over a decade — actually, I would say multiple decades — on the foundations of cognitive computing. That’s how Watson beating Jeopardy actually happened.
And if you look at all the fundamentals here, like machine learning and question answering, knowledge representation, which is at the foundation of cognitive computing, even experiential and interaction modalities and all of those things, those happen through the great work of a whole community of researchers.
And I’m happy to tell you that I have some incredibly bright and accomplished researchers who work with me to make that happen. And that’s the team that created the Watson system that beat Jeopardy. I’m also thrilled to say that we are going to be not only accelerating all of these innovations that go into the new unit that’s announced here today, but we are going to be expanding, and we’re going to be focusing our investment. Almost a third of IBM Research is going to be focused on cognitive computing. And we’re going to be delivering, we’re going to be generating the next generation of all of the foundations and applications and all of the technologies that will keep this going in this very, very wonderful and expansive set of applications that we’ve heard about today from the previous speakers.
So, I want to give you a few examples of the kinds of things we’re doing. And in order to appreciate the examples, I think it would be good to, again, just step back and think about how humans do cognition. When humans do cognition, you first have to sense the environment around you, understand what’s going on. You have to get very good at recognizing patterns of what’s going on around you. And you have to then be able to reason about things that you have seen patterns for, and then you have to be able to go into the sort of the art of it, which is the creative exploration and discovery of it.
Now, I’m going to give you examples of all of those four things that I just mentioned. There’s a number of other things that are foundational to cognitive computing and as we’ve explored in neuroscience and other areas, but I’m going to use those four fundamental faculties to tell you what we’re doing in IBM Research.
So, first, the ability to give Watson the power to see is that of learning from a very large collection of images and multimedia information — videos, audio, animations, if you will — all kinds of information that is not textual. And we’re not just talking about understanding the metadata that’s associated with these images, it’s understanding the content. It’s not only understanding it’s learning from the content over time so when you look at an application like looking for anomalies in an x-ray or in an MRI or any of the other image and video kinds of data sets that we’ve heard about, it really takes a huge amount of expertise on the part of humans to be able to do that. And the accuracy can be greatly improved when you adopt a tool that has learned the anomalies over time through a large number of data sets and through training from human experts.
When you adopt those as assistants, you can improve accuracy, improve productivity, and you can in fact get into much more real-time analysis and diagnosis of many kinds of conditions that we cannot do today. That’s going to be the power of “see.”