The Cavalier Daily
Serving the University Community Since 1890

U.Va. alumnus discusses his design for Sirius

Jason Mars speaks on the future of intelligent personal assistants

At the “Distinguished Alumni Seminar Series” Friday, presented by the computer science department, University alumnus Jason Mars, a recognized global expert in some of today’s most compelling issues in computer science, introduced his design of Sirius, which he described as an “end-to-end [intelligent personal assistant] web-service application that accepts queries in the form of voice and images, and responds with natural language.”

“Computing is changing before our eyes,” Mars said.

As individuals become increasingly interconnected through mobile devices, new technology such as IPAs, including Apple’s Siri, Microsoft’s Cortana and Google’s Google Now, have become more prevalent in day-to-day life. With the invention and heightened popularity of wearable technology such as Google Glass and the Apple Watch, the way individuals best access the internet and make general queries has changed.

“The kinds of apps that we use on our phones [are] living in cloud platform, and it’s been estimated that the data center computing that we have today takes 30 nuclear power plants to power,” Mars said.

Simply, data centers cannot accommodate this emerging class of technology or its possible millions to billions of users. With his research laboratory, Clarity-Lab, Inc., at the University of Michigan, Mars seeks to improve the architecture designs for data centers and data center efficiency at large to ultimately increase usage of IPAs.

An open-source replica of Siri, Sirius’s source code is made available to the general public so that the the program can be altered and improved to produce an optimal product for everyone. Sirius functions through three distinct components — voice command, voice query and voice/imagery query. Through all three components, Sirius is able to respond to commands such as “Call my doctor,” or questions such as “Who is the lead singer of U2?” or “When does this bar close?”

Because of flawed designs in current data centers, there is a large scalability gap, meaning that over 100 times more machines are necessary for users to switch from traditional web searchers to IPAs. In order to narrow the gap, Mars’s research team investigated different types of accelerator platforms, including CPUs, GPUs and FPGAs. Mars’s team discovered that FPGA-accelerated servers had higher efficiency than the GPU-accelerated servers. Efficient accelerators are crucial for the future scalability of IPAs.

Due to the nominal similarity to Siri, Apple sent Mars a letter requesting that he change the name “Sirius.” After consideration, Mars settled on the name “Lucida,” referring to the Latin name for the brightest star in a constellation.

“Lucida hears, sees and understands. … She is a highly intelligent personal assistant that has the ability to complete speech to text, facial recognition, object recognition, question and answer and many more functions,” Mars said.

After graduating from the University with a computer science M.S. and Ph.D. in 2008 and 2012, respectively. Mars currently works at the University of Michigan as an assistant professor.

Clarity-Lab, Inc., an acronym meaning “Cross Layer Architectures and Runtimes in 10 Years,” currently hosts a multitude of Ph.D. students and postdoctoral fellows. The laboratory’s research has resulted in Mars’s publishing of several significant published papers in the field and the reception of several awards, including the prestigious Google Faculty Research Award.

Mary Lou Soffa, Mars’s mentor and professor of sciences in the University’s computer science department, praised Mars’s work.

“He’s only been out just a little over three years, but the achievements he’s produced have been just awesome,” Soffa said. “He’s been extremely productive, and the work he has done has been of the top quality.”

So far, Lucida has had over 10,000 downloads, and its success is ever-growing. For the future, Mars’s philosophy is simple — he aims to observe the trajectory of technology, predict where human desires are headed, determine the best way to reach them and ultimately discover a balance between those desires and human capability.

Comments

Latest Podcast

From her love of Taylor Swift to a late-night Yik Yak post, Olivia Beam describes how Swifties at U.Va. was born. In this week's episode, Olivia details the thin line Swifties at U.Va. successfully walk to share their love of Taylor Swift while also fostering an inclusive and welcoming community.