Next, if you look at how we perceive patterns — again, going back to neuroscience — there are these fundamental blocks of structures in our neocortex which actually hardwire things that we see as patterns around us and other kinds of knowledge that we’ve gained over time.
Now, when you get to the scale of data and when you get to the amount of these patterns that we need to learn and we can learn going forward, our traditional computing architectures do not work anymore. We call those traditional computing architectures Von Neumann machines, these have been around for half a century now, and we’ve all sort of grown up on those architectures.
But we believe and we are proving that in order to get to this new world of huge data and cognitive capabilities, you need a new architecture; we call those architectures neurosynaptic systems. And this is a fundamental rethinking of how computing happens. It actually mimics what happens in the brain through neurons and synapses, and these patterns that I’m talking about are actually the fundamental way in which you specify what these systems do: they learn, they interact, but it all happens because they understand how to work with patterns. And that’s this ability. This is actually a very long-term project that we are doing, and it’s a breakthrough in computer science. And this, we believe will be, will also become mainstream in the future, and we look forward to solving many more applications as we go forward.
You’ve heard a little bit about the power to reason. But I would like to maybe dig a little bit under the covers here and tell you about, when Watson answers a question, it tells you the answer. It tells you what the answer is. But the question of why that is the correct answer is fundamental to many professions. Like the medical profession, a practicing physician needs to understand the logic, the reasons why anyone, whether it’s a human or machine, is giving a particular answer. And they need to analyze it. They need to make a judgment about whether that sequence of logical steps and the evidence underlying that sequence of logical steps are appropriate for the particular situation, because there’s a huge amount of judgment involved here.
So, when you look at the technologies like Watson Paths that was discussed earlier, it gives you the ability to formulate hypotheses or maybe even specify arguments that says, well, what if I did this; would that be supported by any existing literature? Or, what if I wanted to…what if the patient does not want to do something? Or if the patient condition requires that we explore something brand new that we did not know about. What would be the potential options, and what would be the consequence of each one of those options? Those are the reasoning methods that are actually pushing the boundary in not only in our lab but in academia and in the broader scientific community, because those are very, very difficult questions and it really requires deep science to think about.
And when you look at the extreme or the ultimate way of doing this reasoning, it becomes interactive reasoning, because you are having a discussion, even a brainstorming session with somebody, and you’re saying, what if I did this? Would that work? And somebody says, oh, you know, these are the reasons why it might not work. These are the reasons it might work. And you pick up on one of those reasons, and then you start digging deeper and so forth and it becomes an interactive dialogue between a human and a cognitive system that can assist the person to really do what they want to do.
And we’re building systems like that in the lab today. You know, we will have other use cases where we can demonstrate this capability. And I want to get now to the creative portion of what cognitive systems can do. Right? This is about…it’s not about knowing precisely what the question is; it’s about exploring about what the possibilities are, what are the adjacent ideas, what are things that I haven’t even thought about.
What can I discover from things that I know may be plausible but I’m not sure about what the consequences are and what the evidence is. And in the case of examples I use here are those of, let’s say, chemistry, biology, materials science.
You may end up needing a combination, a cocktail of chemicals for a particular medical condition. And we may not be able to imagine what those are without the help of a system that can pull various different pieces of information across the entire space of information and put it all together and suggest something new that we never considered. Same thing with metals.
For specific applications, you may need a combination of elements from the Periodic Table that you may not have considered before, and those kinds of alloys or materials can be discovered through this exploratory brainstorming technique that cognitive computing can enable even human experts to be able to do. That actually brings me to the idea of innovation. Right? Creativity, innovation, we’ve mentioned many, many times that we are going to be opening up this whole platform and the APIs for the broader community to engage and build new cognitive applications.
We call them COGs, by the way, this is just a new terminology that we’re beginning to use. But in order to be innovative, we absolutely need those innovators, those humans, who will be working with these cognitive systems to build all sorts of applications. You know, we of course proved the technology out with our clients. That’s our step one.