Hey, Siri. How Exactly Do You Work?

blog.siriOn Thursday, January 7, the Villanova in the Valley cohort had the privilege to visit Apple Inc. and speak with Bill Stasior, the Vice President of Siri.

With three degrees including a PhD in electrical engineering and computer science from MIT and extensive professional experience in machine learning and search technology with companies such as Oracle, AltaVista, Amazon, and A9, I felt anxious and nervous prior to our meeting; however, Bill’s welcoming nature, humorous personality, and open (as open as an Apple employee can be) discussion with the cohort was educational and insightful.

Bill began our groups’ meeting with a presentation of what defines Siri, recent developments in Siri, and the true impact of Siri.

Beyond helping iPhone users send messages, make phone calls, search the web, and set reminders, Siri is an incredibly complex and nebulous tech, at least to a future lawyer.

However, Bill described Siri as a combination of Speech, Intent, Action, and Dialog.  Speech is turning spoken word into text.  Intent is the process of Siri understanding the written words.  Action is, well action. Dialog is Siri’s response of either “Done” or asking for further clarification.  Whew.

As Bill noted, iPhone users certainly remember the early days of Siri where misunderstanding and errors were commonplace, but major advancements in Siri’s capabilities occurred from 2014 to 2015, such as a word error-rate decrease of 44% in English, 50% in Japanese, and over 60% in Chinese, which has led to the improved functionality and utility of Siri.

Further, Siri has taken on a persona of her own, which Bill’s team worked hard on improving and has caught late night media attention on shows like Jimmy Kimmel and Jimmy Fallon.  Also, Bill showed a clip of how Siri saved the life of a young man, who was involved in a car accident, by alerting rescues services.

After presenting, Bill answered the cohort’s questions involving the future of machine learning, how he uses Siri day-to-day, Apple’s top-down organizational approach, and how the Siri team balances research against development.

For me, Bill’s discussion on the future of machine learning was thought-provoking.  At a basic level, machine learning is a computer’s ability to understand and predict key words, while adapting and personalizing results over time.  Even some basic online research will highlight both the benefits and concerns for the future of machine learning.

As more and more data is created and an increasing number of mobile devices come online, computers’ artificial intelligence (“AI”) increases, which allows for greater ubiquity and incorporation into everyday life; however, as AI increases, some are concerned computers’ intelligence may very well surpass all human knowledge combined.  Currently, as Bill mentioned, computers are capable of helping to perform basic tasks, but computers still lack common sense reasoning.  Thus, AI singularity remains somewhat distant.

Another interesting question posed to Bill was how the Siri team balances research against development.  To start, Bill stated Apple and the Siri team is heavily focused in product development, based off of the numerous products placed into the market, so the real question revolved around research.

While studying at MIT, Bill had over 11 years of research experience.  During our meeting, Bill mentioned the Siri team’s research is as extensive and complex as what he was involved with at MIT; however, the main difference being researchers publish papers, and Apple produces consumer products.  Therefore, research is a vital aspect of Apple’s innovation and continued progress.

In the end, the cohort was fortunate to meet with Bill Stasior, not only due to his expertise and knowledge in Siri, but because of his welcoming and engaging personality.

To learn more about siri click here.