Finding relevant content is tricky business; not only is there an abundance of irrelevant content to confuse us and muddy the waters, but we often simply don’t know something exists until we see it. Traditional search fails us here because that method assumes we know what we are looking for in the first place, or that we possess the forethought to make some assumptions and pick out some key words to enter into the search bar. Perhaps what we really need are intelligent tools that suggest things to us before we even know we are looking for them.
TrapIt, a new online content discovery tool, aims to meet that need. TrapIt was developed by the same minds that created and then sold Siri to Apple (the Siri technology has now been fully integrated into the iPhone 4S). Both offerings leverage artificial intelligence (AI) technology that was developed as part of the CALO (Cognitive Agent that Learns and Organizes) project, an AI project funded by the Defense Advanced Research Projects Agency (DARPA). Originally intended for military use, the cognitive software was designed to learn from experience, take orders, explain its own actions, and respond to unexpected input.
The key difference between the two is that Siri is a virtual assistant that responds to your verbal commands to do tasks such as setting reminders or searching for information while TrapIt is designed to seek out the things that it has determined you will find relevant, without you having to ask for them.
TrapIt is currently in beta and delivers personalized content from over 100,000 sources based on “traps”, which are essentially search terms that the user sets up. Once created, a trap will automatically refresh itself with new content to be read when the user logs in, building a personalized homepage that reflect the user’s interests. Creating the trap “tablet usage” for instance creates a stream of content that relates to tablets and usage data. The user is then prompted to give thumbs up or down to the content to indicate whether it is what he was looking for or not. When giving a thumbs down, there are options to indicate why, such as because it was not interesting, the source was not trusted, or the content was spam. This helps to further refine the content that is suggested going forward.
The content that TrapIt collects is refined as TrapIt analyzes how often the user clicks on specific types of content, as well as through the thumbs up or down mechanism. Because the AI learns as TrapIt is used, it is too early to tell from my tests how effective it will be at providing relevant content, but results so far are encouraging.
Siri, the first commercial application of the CALO AI, has been well received and is quickly becoming a popular feature of the iPhone (as well as being hacked to run on older iPhone models and platforms such as Android, or even to control thermostats). Now that TrapIt is applying the same underlying technology to content discovery, we will have a chance to see how effective the AI really is, and if it can recommend content in a way that helps to cut through the clutter of information and get us the information we really need.
Cody Burke is a senior analyst at Basex. He can be reached at firstname.lastname@example.org