Companies that know how to harness machine learning and artificial intelligence are clear winners in the 21st century economy. And, arguably, few companies on Earth rival Google for its ingenuity in unlocking their powers. Suzanne Frey, SF ’06, Director of Security, Trust, and Privacy at Google, says, in fact, that every area in which Google is advancing relies on the progress of AI and ML.
DeepMind
Frey says that a game-changer was Google’s 2014 acquisition of the London-based AI pioneer DeepMind. Now a world leader in artificial intelligence research, DeepMind is developing programs that can learn to solve complex problems without necessarily needing to be taught how. Working closely with subject experts, DeepMind researchers are building tools that augment advances in fields such as healthcare and energy. And, of course, it’s tapping its late-breaking knowledge to enhance Google’s own products and services.
Frey says that DeepMind’s capabilities were best demonstrated recently on the virtual playing field. “The work we’ve done with DeepMind really went big when we won a pivotal GO match against the #1 champion in the world, an accomplishment that most thought was a decade out on the computer science horizon.” Big deal…just a game? Frey explains why not. “In winning that match, DeepMind illustrated the ability to run an exceptional number of deeply complex scenarios in parallel, at scale, and select the best outcome. And that capability is applicable to so many domains—the biosciences, in particular, allowing biotech researchers to do drug testing and create genetic scenarios at scale without testing on animals. Bottom line: the implications of that winning game of GO could have enormous positive impacts on society.”
Productivity apps
As a time-challenged tech executive, Frey is understandably somewhat obsessed with time-saving apps. “My twin passions for AI and time collide heavily at Google. Anything that saves time and optimizes energy is important. Google, she says, is using AI to create a suite of indispensable apps like Smart Reply in Gmail, which offers you quick responses to choose from so that you don’t have to compose a message from scratch. Another app will note a meeting on your calendar and queue up related documents in preparation. All these little inventions, Frey says, can save a minute here, five minutes there, minutes that add up by the end of a given day. “Nothing is more precious than time; just think about the total area under the curve saved by each of these features, and I’m excited to say that there’s much more time savings to come.”
Voice-activated searches
On the educational horizon, Frey reports that Google’s search functions have advanced to such an extent that a Google search is almost akin to having a research assistant. Google Scholar helps students find relevant resources across the world of academic research, pulling in articles, theses, books, and abstracts from publishers, universities, professional societies, and other online repositories. Google Scholar ranks documents the way researchers do, weighing the full text of each document, where it was published, who it was written by, as well as how often and how recently it has been cited in other scholarly literature. “The quality of the search,” Frey points out, “is directly correlated to the quality of the AI.”
Google Home Assistant
“Hey, Google. Is Johnny’s Pie in the Sky open? Order me a pizza with pepperoni and mushrooms. And tell me how to get there. Is it on Front Street?” Frey gives this example of a series of tasks that Google Home Assistant could handle for someone returning home from a stressful workday. Again, she says, it’s about conserving that most valuable of all commodities—time. And as the AI becomes more sophisticated, so does the Home Assistant. “One of the most exciting developments in voice-assisted searches is that the ‘trail of thought’ problem has been solved,” Frey says. “We’re teaching AI to realize that your prior query is related to your next query. It understands the pronouns within a string of questions. After you asked if the restaurant was open and you followed it with the phrase ‘How do I get there?’ the Home Assistant knows that you’re asking how to get to the restaurant, even though you haven’t specifically used those words.”
The Home Assistant, Frey says, is an illustration of Google’s intention to reposition itself as a personal assistant rather than a comparatively aloof search function. “There is no part of Google that isn’t directly investing in machine learning and AI,” Frey says. “It’s all connected to our mission to move from searching to assisting. It’s a thrilling space to be working in right now, even in security and data protection, and I’m enjoying every minute of it.
The post Deep mind thinking about AI and ML appeared first on Leadership Blog.