Tse: How is the new, Watson-created, personal shopper different from the other automated queries in existence?
Gold: Watson is really a very different type of technology. It's really a new technology that affords an individual the opportunity to ask questions in natural language. So rather than a programmatic input, I can express my customer query as I would a friend, and it then has the unique ability to really rediscover in context information that's been made available. So if you're thinking about maybe asking a question about a product or seeking reviews of a product at that very moment, you're able to direct that question through the service.
The service investigates all the information that's been populated and then it brings back. But what's really interesting, what I think is extremely interesting is that it does so in a very transparent way. It brings back not just a single response -- it can bring back a set of responses to a request and can weigh those responses with confidence, as to whether they'd been an appropriate type of response to the question. It also brings back all the supporting evidence. So unlike all the other forms of information discovery where I have to rely on the system to make a determination for me, this is actually an intelligent system that's capable of understanding much the way you and I would have a conversation.
We now can have a conversation with Watson.
Tse: Are the queries to be made with voice or typed commands?
Gold: Well it's agnostic in the sense that really, the technology can process a natural language input. That could be from a text, that could from something that was keyed in, it could be through a technology that did a voice of text translation. That's out there so we can support those technologies. But the essence of it is that the input itself is unstructured right, so it's in our native language. And what's interesting about that is that it's not voice recognition.
We've all been on those voice recognition systems, pounding on the zero key yelling "agent, agent, agent."
This Watson technology is very different in the sense that it actually understands the nuances -- the colloquialisms, the idiosyncrasies of the human language.
We don't really think about the way we talk to each other as illogical, but at least in a historical frame of reference to computerization it is, right. In English, we say things like "houses burn up as they burn down," right. And that's counterintuitive. And you know we say things like "noses run and feet smell." You look at a classical definition and you wouldn't uphold that.
So when I talk about this natural language capability, it's really in and of itself quite an accomplishment. And that's a part of Watson. But its only one dimension, right. That by itself is interesting, but then to literally be able to read context ...
We live in a world where there's an information deluge right. 90% of the world's data was created in the last two years, 80% of that's unstructured, and it's not just blogs and tweets and posts, it's information that's been captured and stored away in recordings, surveys and forums. I mean how often do I call a vendor and hear, "this call is being recorded for quality purposes." I can't help but wonder "what do they do with that recording?"
Imagine now if you transcribe it, and literally a system can learn from that experience, can learn from that interaction with the consumer and get smarter -- which is the third part of Watson. It learns -- so unlike a typical conventional system that we know today that's programmatic in nature since the '50s, that's what we've known in computing, it's programmatic systems. They're driven by logic, they're driven by rules based on structured data. Now remove those barriers and imagine a system that learns. It gets smarter with each interaction, with each outcome. With each new piece of information, it's getting progressively more insightful and I always think about this in the context of today.
If I use a piece of software and it doesn't work as designed or it doesn't have the capability that would be beneficial, I would have to wait until the next version or the next release. And then it may or may not be addressed. With Watson, it's progressively advancing every time there's an interaction. And again, it's learning based on behavior, based on activity, and it turns out that that's a very important distinction as we think about how we as individuals live our lives.
Imagine computers have historically been deterministic right, they've always been looking for the single right answer. Watson has the ability to take in that new data and say, "you know what, I know yesterday I was 90% confident that the world was flat, but today I'm only 84% confident because there's new evidence that would controvert that statement." With further explanation and further use, the system gets smarter right and eventually learns that the world is round. And this happens every single day. In healthcare, finance, law, retail , every industry, things are changing. And we want to be able to take advantage and learn from that.