Baidu is powerful new contender in deep learning race May 20 2014

Chinese search giant sets up Silicon Valley lab as it fights with Google, Microsoft and others to redefine mobile search

By Caroline Gabriel

Behind nearly all the interesting new mobile web business models lies a massive artificial intelligence, or deep learning, engine. Some of these are being developed in great secrecy to support the next wave of services from Google, Facebook or Microsoft. Others, like IBM's Watson and AlchemyAPI, are open to a broad developer community, while the next stage will be 'as a service' platforms - a possible business for some operators, like AT&T, which have developed advanced analytics systems to examine their own networks and subscriber behavior.

But don't be fooled by the fact that all these names are American. As in most areas of advanced technology, Chinese web giants and research institutions are matching their western counterparts dollar for dollar in development of next generation AI engines, with a particular focus on supporting mobile services such as advanced context-aware search (of which Apple Siri, Google Now and Microsoft Cortana are just infant examples); as well as analytics and personalization services for enterprises and IoT (internet of things) providers.

The 'Chinese Google', search provider Baidu, is particularly active in developing mobile and web platforms which will start by dominating its home market, but will also look for global presence, especially in emerging markets like India and Indonesia, where no player is yet as entrenched as Google is in the west. As well as their substantial homegrown R&D resources, the big Chinese players are going global in terms of tapping expertise too.

Baidu has set up a research lab in the heart of Silicon Valley, in Yahoo's home town of Sunnyvale. And it has hired a Stanford computer science professor, Andrew Ng, to head the global Baidu Research operation. Ng is a catch - he worked on deep learning at Google when it was doing its 'Google Brain' project; was a founder of online learning organization Coursera; and is a well-known figure in the AI field. He illustrates how Silicon Valley brains are no longer a resource mainly available to US firms. While based in Sunnyvale, where Baidu says it will invest $300m in its new facility over five years, he will also oversee several labs in China.

Baidu also set up its Beijing Deep Learning Lab last year and says it has already made progress in key technologies which will revolutionize the online search experience, making it open to many inputs (such as gestures and voice) and able to predict users' needs and behaviour, based on sophisticated analysis of context, location and history. That prediction can then be the basis of all kinds of analytics to support activities like targeted advertising, demographics research (as well as more sinister 'big brother'-type uses). Among Baidu's areas of research are image recognition and image-based search, voice recognition, natural language processing and semantic intelligence, machine translation and advertising matching. The firm has already introduced an app to identify objects in smartphone photos.

Kai Yu, director of the Deep Learning Lab, told MIT Technology Review that the new Silicon Valley lab would be targeted with fundamental research, while his lab will examine how to apply the deep learning breakthroughs to new commercial products and services.

Google will be eyeing these developments with fear. It is essential to its power and revenues that it continues to dominate and shape the search experience and just last week, CEO Larry Page acknowledged in his annual 'Founder's Letter' that "in many ways, we're a million miles away from creating the search engine of my dreams - the one that gets you just the right information at the exact moment you need it with almost no effort".

"Information is Google's core," as Page noted in his letter - gathering and analyzing vast quantities in all kinds of formats, in order to deliver the new breed of search results, but also to track people and machines more efficiently than ever before, to power the core revenue generators, advertising and big data. As well as Google Brain and the Google Now engine, it acquired DeepMind to bolster its capabilities.

However, others want to usurp its position. Microsoft recently set up a new special projects group within its R&D organization, according to insiders, which aims to take on the similar Google X unit, and is headed by Norman Whitaker, formerly deputy director of innovation at the US Defense Department's research arm DARPA. The new Microsoft initiative is working on "disruptive technologies that could benefit the company and society" and has a long list of targets, including machine learning/AI, mobility, big data, distributed computing and user experience design, according to a recent job advert.