Research teases capabilities of the next-generation of Siris and Cortanas

Researchers at MIT have developed software that could take virtual assistants a step closer to replicating the abilities of a real-life PA.

An early demonstration of Siri in 2011
Juggling appointments can be a nightmare and virtual assistants are far less helpful at managing your schedule than their real-life counterparts.

But just as a person might suggest holding a lunch meeting in the nearby Hilton to give you more time to get to your 2pm conference call, so the successors to Siri and Cortana might be equally adept at managing our lives.

Researchers at MIT are working on planning software that could help virtual assistants work out the best outcomes based on the constraints you give them. For instance, if you need to grab dinner but can't be out for longer than four hours it might recommend going to Applebees rather than TGI Fridays. If you're adamant about your meal choice it could just let you know the probability of getting back within four hours for each diner.

At the annual meeting of the Association for the Advancement of Artificial Intelligence (AAAI) this month, the researchers will present algorithms that represents a step towards what they describe as "a better Siri".

The application being worked on by the group at MIT's Computer Science and Artificial Intelligence Laboratory has the same underlying framework used by software that NASA has used to plan missions.

To demonstrate how planner might work, the team references another potential application for the software, scheduling bus routes. In this instance the software could be given the constraint that buses on a route should reach their destination at 10-minute intervals - and reliability thresholds, such as buses should be on time at least 90 percent of the time.

Then, on the basis of probabilistic models - which reveal data such as that travel time along this mile of road fluctuates between two and 10 minutes - the system determines whether a solution exists. For example, perhaps the buses' departures should be staggered by six minutes at some times of day, 12 minutes at others.

If, however, a solution doesn't exist, the software doesn't give up. Instead, it suggests ways in which the planner might relax the problem constraints: Could the buses reach their destinations at 12-minute intervals? If the planner rejects the proposed amendment, the software offers an alternative: Could another bus be added to the route?

This additional insight the planner offers come at a cost, however. The need to calculate probabilities, for instance of the time it takes to travel a mile of a bus route - adds a large computational overhead.

"It's always hard working directly with probabilities, because they always add complexity to your computations," said Cheng Fang, a graduate student in MIT's Department of Aeronautics and Astronautics that worked on the software.

"So we added this idea of risk allocation. We say, 'What's your budget of risk for this entire mission? Let's divide that up and use it as a resource.'"

Telling the system the planner can tolerate a certain amount of failure allows it to discount portions of the calculations, which makes the calculations simpler to resolve.

Two separate papers will be presented at the AAAI focusing on calculating solutions for the planner and identifying constraints that would prevent a problem being solved.

Both procedures rely on graph theory. In graph theory data can be represented as nodes, usually depicted as circles, and edges, usually depicted as lines connecting the nodes. Any scheduling problem can be represented as a graph. Nodes represent events, and the edges indicate the sequence in which events must occur. Each edge also has an associated weight, indicating the cost of progressing from one event to the next - the time it takes a bus to travel between stops, for instance.

The MIT researchers' system identifies which problems are insoluble and then calculates the lowest-cost way of rebalancing the loop, which it presents to the planner as a modification of the problem's initial constraints.

"These papers are quite interesting," said Jiaying Shen, a research scientist at Nuance Communications, which is widely believed to have developed the voice-recognition technology used by Apple's Siri. "They had a flurry of papers on chance constraints, but in the recent papers, they added uncertainty in there, which makes the problems that it can model more complicated and unpredictable and more realistic."

Nuance is also working on technology to boost the capabilities of virtual assistants, by plugging them into data that allows them to understand more about the world and your life - your likes and dislikes, where you are, what you are doing at a moment in time.

At the forefront of this work is Project Wintermute, which shares a name with the powerful AI from the classic cyberpunk novel Neuromancer.

About Nick Heath

Nick Heath is chief reporter for TechRepublic. He writes about the technology that IT decision makers need to know about, and the latest happenings in the European tech scene.

Editor's Picks

Free Newsletters, In your Inbox