Today in APIs: Sofilia Promises Computer Conversation Capabilities

The creator of Sofilia wants to blow away other digital assistants with real conversation. A hackathon focused on dementia creates solutions based on a design principle everyone can learn from. Plus: Huawei launches new video IP solution, and Jetstream enhances the ConnectWise Integration.

Sofilia Aims for Meaningful Conversation Between People and Computers

You don't have to use Siri or Cortana for long before you understand the limits of how computers respond to human. Today's digital personal assistants can easily move from novelty to annoyance. Yesterday, we discussed the race to commoditize these services with an API that can easily bring this technology to any app. The need for an API of this nature heralds the ubiquity of convenience. But in the end, talking to Siri feels oddly similar to using a mouse or multitouch pad. These tools are valuable but remain constant reminders that the days of real human/machine discussion of the sort we read about in scifi novels feels as far in the future today as as it did when the first stories started appearing 75 years ago.

Dave Grundgeiger aims to leap past all that to create meaningful conversations between people and computers right now. He points out that all the fancy Siri-like tools work by extracting definitions from text. To get to meaningful conversation, he says, the software will also need to create a model of the world and React to what you say based on that evolving model. He has framed his goals for Sofilia in a way that any Siri user can appreciate. It will understand what you are asking for better, clarify requests and engage in conversation to learn about you, recall and integrate everything you tell it, and build a model of the world and of itself over a lifetime of use. This means that each installation of Sofilia will become unique to the person using it. His approach has a quintessentially quixotic feel to it, and he's launched a kickstarted campaign that is underway right now to raise what can only be described as a paltry sum of money next to the height of the bar he is aiming to clear: just $75,000. That's right; he's aiming to crack what would make it possible for a computer to act like the one in the movie Her, for example, for that kind of money. To be fair, he says he has been working on this for over a decade. So perhaps this is just the home stretch, which might explain the tiny sum.


As Grundgeiger explains, he's taking a unique path, whose benchmark is Microsoft's Machine Comprehension Test:

Sofilia is software that uses an augmented grammar to convert natural-language text to and from a "semantic triple" database, or "triplestore." Semantic triples are a way for computers to store, manipulate, and reason with meaning. There is a lot of software today for working with triplestores in various contexts, but you need a person to create the semantic triples manually or by using special "semi-structured" languages. There is currently no software that converts unrestricted natural-language text to and from semantic triples. Again, that's what Sofilia is for.

Is Grundgeiger a crackpot or is he about to blow the hinges off computing as we know it? If he makes his promised delivery date, we won't have to wait too long to find out. He estimates arrival of Sofilia in November of 2015.

A Hackathon Focused on Dementia Creates Surprising Wins

Dementia is like falling off a cliff: once someone has toppled over the edge, the trajectory is one of sad inevitability. Why focus a hackathon on trying to affect the inevitable? Dmentiahack, held at Ryerson’s Digital Media Zone Sept. 12-14, yielded some surprising answers. But before we look at those, Calvin Dao and Andrea Vacl report in The Eye Opener, the most powerful answer as to why a hackathon was worthwhile comes from those who suffer themselves:

Brenda Hounam, who suffers from dementia and spoke at the event on Saturday, expressed her need for technology and described how game apps on her iPad “help clear the cloudiness in [her] mind.” The technological entries at the event were judged on ease of use, consumer interest and potential business opportunity.

The first prize went to CareUmbrella, an app that uses labels that can be stuck to things that then use NFC to trigger reminders and information on digital devices. For example, tapping a phone on a sticker that is on a microwave can bring up an instructional video about how to use it. Nitin Malik received a runner up prize for his team's innovation, "All the Pi." It creates audio reminders to help patients complete tasks such as remembering to turn the light out after it was turned on earlier. It also sends push notifications to caregivers when tasks go uncompleted. Many hackers talked extensively with people who have dementia to come up with and modify ideas. Beyond the importance of compassion, there's a critical design principle at work applicable to APIs, programming and hardware of all sorts: run things by users for feedback.

API News You Shouldn't Miss

Be sure to read the next Hacking article: Today in APIs: Million Dollar Hackathon Returns