NEW YORK—When IBM opened up its Watson cognitive computing platform to developers last November, it launched an effort to enable anybody with programming skills and an idea to take the technology and use it as a foundation for building apps that can both understand natural language and learn new things to expand its knowledge base.
Watson represents a breakthrough in the field of artificial intelligence, and IBM's opening up the platform indicates just how far AI has come. At the GigaOm Structure Data conference here, IBM and AlchemyAPI spoke about their experiences in providing cognitive computing technology to help solve broad societal, medical and other problems, and also to help democratize the use of cognitive technology. Indeed, AlchemyAPI says its goal is to democratize breakthroughs in deep learning to power unstructured data applications.
Elliot Turner, CEO of AlchemyAPI, and Stephen Gold, vice president of worldwide marketing and sales operations for IBM's Watson Group, spoke on the issue of democratizing artificial intelligence with APIs.
"We're in this new trend—the API economy where you take something complicated and expose it as an API," Turner said. "Cognition is yet another one of those things. "So we're seeing companies leverage unstructured data—things like photographs, videos, chat logs, documents—to make better, more informed business decisions to automate processes. They're leveraging humanlike capabilities inside automated workflows. I think these technologies can ultimately augment what's possible in business and humanity, but not necessarily replace that."
As a society, we are trying to wrap our mind around exactly what is cognition, Gold said.
"Even though we think about a humanistic aspect of computing, when we look at it, we're very tied to this notion of logic and rules and reason and structure," Gold said. "And the vast form of data that's proliferating today that's available to us is all unstructured. It's the texts, the blogs, the tweets, the articles, the photographs.
We're just now starting to experience the idea of a system that is readily approachable that you can navigate through natural language. And better yet, these cognitive systems are systems that learn. So they get progressively smarter.
Gold noted that, in medicine today, less than half the discipline is evidence-based. "This holds the promise to take us to a point that our physician will have an assistant that can actually provide, unequivocally, insights that were locked away in some study or trial or periodical.
Turner acknowledged that data and learning is a lot more nuanced in cognitive systems. "The whole idea of a ground truth, which used to be a big thing in computing, is really starting to go away," he said. "Our systems leverage a technique known as unsupervised learning—where, by exposure to data over time, you can build up a conceptual mapping of the world. In terms of keeping a system from going off the rails, one of the things that we saw in exposing our system to the Internet is that dogs are people. So the point about context being key is really important. You can have multiple ground truths, depending on the situation."