After losing a pioneering leader, where is Google's AI search going?
Google's ubiquitous search engine could be getting even smarter, despite the retirement of the man behind its rise.
Amit Singhal, the longstanding chief of Google Search operations, announced his retirement Wednesday effective later this month. Although the loss of Mr. Singhal after 15 years with the company is significant, Google鈥檚 choice for his replacement shows the direction in which it hopes to take its Search product.
鈥淪earch is stronger than ever, and will only get better in the hands of an outstanding set of senior leaders who are already running the show day-to-day,鈥� . 鈥淪earch has transformed people鈥檚 lives; over a billion people rely on us. Our mission of empowering people with information and the impact it has had on this world cannot be overstated.鈥�
One of the senior leaders Singhal referred to, John Giannandrea, will be after Singhal鈥檚 official departure. Mr. Giannandrea is currently one of Google鈥檚 vice presidents of engineering and has worked at the company since 2010. But the appointment of Giannandrea, a leader in the development of machine learning at Google, demonstrates how the company hopes to incorporate artificial intelligence (AI) into one of its most recognizable products.
Machine learning refers to a field of study related to AI that focuses on how computers adapt to evolving problems and "learn," based on real-world information and guidance. This process results in more organic choices or predictions by machines that would not be possible using only programs written by human coders, and is one of the uses of AI that Google is investing its resources in.
鈥淢achine intelligence is crucial to our Search vision of building a truly intelligent assistant that connects our users to information and actions in the real world,鈥� Google told Reuters in an email.
Google has already begun integrating AI into its services, including Search. The company announced in the fall that RankBrain, an AI system that can understand, filter, and connect words and phrases 鈥� even those it doesn鈥檛 yet fully know or understand 鈥� is of all new Google Search queries.
鈥淢achine learning isn鈥檛 just a magic syrup that you pour onto a problem and it makes it better,鈥� Google research scientist Greg Corrado told Bloomberg. 鈥淚t took a lot of thought and care in order to build something that we really thought was worth doing.鈥�
Both Google users and programmers agree on the process' worth. The company has found RankBrain to be better at ranking Web pages for searches than humans; the AI system scored 10 percent higher on search evaluations than Google鈥檚 own engineers.
While RankBrain was initially approved by Singhal, he was hesitant to more widely incorporate machine learning technology into Google Search due to the potential unpredictability and lack of transparency in how the AI actually functions.
鈥淧eople understand the linear algebra behind deep learning. But . They鈥檙e machine-readable,鈥� Chris Nicholson, the founder of the deep learning startup Skymind, told Wired. 鈥淭hey can retrieve very accurate results, but we can鈥檛 always explain, on an individual basis, what led them to those accurate results.鈥�
Whereas adjustments to the programming behind Google Search could previously be made by engineers through changes in algorithms that were written by humans, modifying an AI-enhanced Search would have to be done through 鈥渋ntuition, trial, and error,鈥� according to Mr. Nicholson, as the technology is designed to adapt and change from its original state.
Google鈥檚 advances in AI and specifically machine learning are not limited to Search. for its email service, an聽adaptive computer board game strategist, self-driving automobiles, and more, are all in the works at the technology company.
鈥淢achine learning is a core, transformative way by which we鈥檙e rethinking everything we鈥檙e doing,鈥� Google CEO Sundar Pichai said in October. 鈥淲e鈥檙e thoughtfully applying it across all our products, be it search, ads, YouTube, or Play ... We鈥檙e in the early days, but you鈥檒l see us in a systematic way think about how we can apply machine learning to all these areas.鈥�
