Building on its promise to give the entrenched a run for their money, conversational AI startup Ozlo is making its meticulously crafted knowledge layer available for purchase today. Ozlo’s new suite of APIs that includes tools for both expressing knowledge and understanding language will help to democratize the creation of conversational AI assistants.
In the spirit of the expert systems of the 1980’s, knowledge graphs are about leveraging massive fact databases to build intelligent AI tools. The problem is that intelligence isn’t just a matter of knowing facts. Though no truly intelligent AI has emerged from well organized information, Google, Microsoft and others have made use of the knowledge graph to reduce the barriers to getting desired information — think of the search widgets that tell you Barack Obama’s age so you don’t have to click on Wikipedia.
But while the knowledge graph has been usurped by natural language understanding as the most prominent part of the typical conversational AI tech stack, it remains a critical layer. Part of the reason for this is that it has become difficult to compete on knowledge because there aren’t many tools available and developers who choose to start from scratch waste precious time reinventing the wheel.
Ozlo’s probabilistic approach to the knowledge graph simultaneously attempts to make measured technological gains and push the market for knowledge closer to democratization. The tools being released today for developers power Ozlo’s own bot. Ozlo knows over two billion facts, but it can also handle queries for opinionated information — like whether a given restaurant is good for oysters, à la the image below.
With greater clarity than ever, Charles Jolley, Ozlo’s CEO, is pushing the company to provide the enabling infrastructure for conversational AI. Ozlo launched its own consumer assistant back in October, but it doesn’t seem like it will be a major part of the company’s growth strategy. Instead, Jolley hopes that Ozlo can monetize its APIs to open up the playing field for companies to build better experiences for their users.
“The important question is how you tie together the data set with machine intelligence to just create a good user experience,” explained John Lilly, partner at Greylock and investor in Ozlo.
The knowledge graph or Data API is being priced by volume. But additionally, Ozlo is making an Intent API available to help developers identify user intent from queries. A third, Converse API, works to preemptively offer responses that mesh with where a conversation is likely headed.