- The research initiative will involve collaborations with people in multiple Google product groups, as well as professors from Harvard and MIT.
- More informative explanations of recommendations could result from the research over time.
Alphabet on Monday said it has kicked off a new research initiative aimed at improving human interaction with artificial intelligence systems.
The People + AI Research (PAIR) program currently encompasses a dozen people who will collaborate with Googlers in various product groups — as well as outsiders like Harvard University professor Brendan Meade and Massachusetts Institute of Technology professor Hal Abelson.
The research could eventually lead to refinements in the interfaces of the smarter components of some of the world's most popular apps. And Google's efforts here could inspire other companies to adjust their software, too.
"One of the things we're going to be looking into is this notion of explanation — what might be a useful on-time, on-demand explanation about why a recommendation system did something it did," Google Brain senior staff research scientist Fernanda Viegas told CNBC in an interview.
The PAIR program takes inspiration from the concept of design thinking, which highly prioritizes the needs of people who will use the products being developed.
While end users — such as YouTube's 1.5 billion monthly users — can be the target of that, the research is also meant to improve the experience of working with AI systems for AI researchers, software engineers and domain experts as well, Google Brain senior staff research scientist Martin Wattenberg told CNBC.
The new initiative fits in well with Google's increasing focus on AI. Google CEO Sundar Pichai has repeatedly said the world is transitioning from being mobile-first to AI-first, and the company has been taking many steps around that thesis. Recently, for example, Google formed a venture capital group to invest in AI start-ups.
Meanwhile Amazon, Apple, Facebook and Microsoft have been active in AI in the past few years as well.
As for design, that isn't new to Google, either. The company implemented a redesign for several of its apps in 2011 and in more recent years has been sprucing up many of its properties with its material design principles. And in 2016 John Maeda, then the design partner at Kleiner Perkins Caufield & Byers, pointed out in his annual report on design in technology that Google had been perceived as improving the most in design.
What is new is that Googlers are trying to figure out how to improve design specifically for AI components. And that's important because AI is used in a whole lot of places around Google apps, even if you might not always realize it. Video recommendations in YouTube, translations in Google Translate, article suggestions in the Google mobile app and, yes, even Google search results are all enhanced with AI.
