Prior Knowledge debuted Veritable, a predictive database for application developers at Disrupt San Francisco 2012 today.
Based upon work first done at MIT, the service is offered through the company’s Veritable API, which the company says is designed to make applications smarter. The technology uses state of the art machine learning behind the scenes to handle the complexity for real world data.
It’s obvious that we are in the data age. There is plenty of data – too much for us to use, really. But if data is plentiful, then the people who know how to use it are scarce.
“There are just not enough of them,” said CEO Eric Jonas about the data science community. “McKinsey estimates we will be 200,000 short in next few years. There are only so many super nerds out there.”
Using data is a messy task. It’s an issue of semantics. When thinking of a person – how do you quantify that person’s data? How do you measure their height – by inches or by centimeters?
The Prior Knowledge team have spent years learning the art of statistics. What they say they have done is build that knowledge into the software. The knowledge built into the Veritable API allows developers to build apps that can determine what something is by looking at it as a whole. It does this by looking at the causal relationships in data. For example, Prior Knowledge worked with a retailer to determine customer purchasing patterns.
The service helps “magically fill in values,” that may be missing, Jonas said in his presentation today. For instance, a dating site may be missing values about the people that have signed up for the service. “What we are really doing is disrupting blanks.”
The end result is a service that helps people see information that is associated with the data they already have. It’s that information that people may not know. Prior Knowledge is banking that information will be valuable to developers.
The added benefit is in how much Prior Knowledge has abstracted the complexity of doing such complex analysis. You don’t need to have a deep understanding of linear regression analysis. Instead, developers need to know the basics of traditional databases to use the service.
Here’s an example of how Prior Knowledge can use the service to help predict customer behavior:
Jonas says the Prior Knowledge focus is pretty universal.
“We are opening to the world,” he said. “We are trying to get as many possible people across different verticals.”
Prior Knowledge is one of those technologies that can do so much that it is hard to find a place to start. That’s the challenge the company faces in the overall market, especially as it competes with business intelligence companies that are also developing predictive analytics technology.
Prior Knowledge makes insight and predictions as easy as using a database, by building infer-structure – infrastructure for inference. Veritable, their predictive database, goes beyond traditional big data analytics by learning the deep structure of data and generating predictions that reflect all of these relationships. The complex probabilistic models behind PK’s infer-structure grew out of many years of research at MIT. Prior Knowledge continues to advance the state-of-the-art in probabilistic inference in order to tackle the size and complexity of...