So why did Google acquire Kaggle?

Google’s heft in artificial intelligence and machine learning is on full display at the ongoing Google Cloud Next 2017. Reasonably so, while Google is a late entrant to the cloud market, trailing behind market leader AWS (launched in 2003) and Microsoft Azure, it has significantly ramped up its Google Cloud ML engine that was introduced last year in beta stage. Making the announcement at the keynote address was Fei-Fei Li, Chief Scientist at Google Cloud AI and Machine Learning, also the director of the Stanford Artificial Intelligence Lab and the Stanford Vision Lab. At the keynote address, opened by Diane Greene, Senior Vice President of Google’s cloud business and founder of VMWare, the shift to cloud was emphasized by the impressive roll-out of legacy clients – HSBC, Home Depot, Verizon, ebay, SAP and Colgate Palmolive. However, it was the transformational power of AI combined with Google’s Cloud phenomenal force of computing that was the recurring motif.  As Li, a computer vision and AI technologist revealed in her keynote speech, “Google’s Cloud Platform already delivers our customer’s applications to over a billion users every day. And that’s a lot of participation, now if you can only imagine combining the massive reach of this platform with the power of AI, making it available to everyone. From manufacturing, healthcare, finance, retail to agriculture, AI has the power to touch all aspects of life, this is why delivering AI and machine learning through Google Cloud excites me”. And it is the search giant’s bid to lead AI race that saw it snapping up Kaggle. To Google, Kaggle – the largest data community in the world fits snugly in their democratization puzzle. “Artificial Intelligence requires huge amount of data to develop insights and datasets are among the steepest barriers to overcome. It is our attempt to make data available to machine learning experts, developers and eventually our business,” she said. Kaggle team has been building an unprecedented community by hosting contents and making new datasets publicly available. By merging with Google Cloud platform, Kaggle community gets direct access to the most advanced machine learning environment as well as provides a direct path to market their models. “Kaggle has the largest concentration of machine learning talent in the world and speaking of talent and expertise we are deeply committed to help our partners and customers to develop more machine learning and artificial intelligence expertise,” she shared.

Google Cloud democratizing AI

According to Li, AI remains a field of high barriers and it requires rare expertise and resources few companies can afford on their own. “That’s why cloud is the ideal platform for AI and it is also why we are making huge investment in cloud AI and ML, that will emerge in the next years in the form of powerful, easy to use tools that will give every cloud customers entry and into this field . In other words, Google Cloud is democratizing AI, this entails four broad steps – democratizing computing, democratizing algorithms, democratizing data and talent and expertise,” she elaborated. According to Li, first and foremost Artificial Intelligence requires enormous computing. In today’s time, a deep learning algorithm can easily boast of tens of millions of parameters and billions of connections; training and using such models requires computational resource. “That’s exactly what the Google Cloud was designed to deliver last year,” she noted. Li elaborates how AI can revolutionize enterprises and solve real world problems in real-time

• The Google Cloud ML engine is a platform that can harness all the computing power and deliver it transparently. Simply put, one can deliver machine learning models however they like, using familiar tools like the Tensorflow library.
• Machine Learning engine allows one to focus on the creativity of their solution and leave the infrastructure part to Google. When it is time to train the models, upload them to the cloud and allow ML engine that can work at much faster and at much larger scale
• Finally, deploy the results anywhere from your own premise to a mobile device where it can compute its training to use and solve real world problems.

Google APIs – easiest way to put AI to work

However, even with all computing power in the world, AI is the most complex field in computer science. And developers are not quite ready to build their own models. Li says the easiest way to put AI to use today is through one of the APIs Google has provided to deliver fully trained machine learning models and tackle common problems. “These APIs are like a switch that immediately activate an intelligent component in any application, allowing it to understand speech, photos, or translate text or parse natural language, but there’s a lot more in Google’s depth and breadth of AI technology,” she said. Vision API: Vision API had been under steady development and featured some significant new capabilities. It is an expansion of the APIs metadata to recognize millions of entities from Google’s knowledge graph for images that are available on the web. Infact, the same metadata powers Google’s image search. The second phenomenal feature is an enhanced optical character recognition, capable of retrieving text from images from text heavy documents such as legal contracts and other complex papers. Video Intelligence API: The world of pixels goes beyond pictures and understanding the rich content of video has been a tremendous technology challenge for many years. Computer vision researchers have often considered videos the dark matter of digital universe, she divulged, unveiling an entirely new API powered by machine intelligence – that tells at high level what the video is about and at a granular level what labels define each video scene. In enterprise settings, Video Intelligence API can be used by media publishers to harness important information embedded in videos.

AI at work @Google and use cases

Citing examples of how Google leverages AI, Li revealed Youtube’s recommendation list are powered through AI.  Machine Learning algorithms are already helping Google’s ad sense in shopping to deliver relevant information to our customers. Talking about the appeal of self-driving cars, an area of high interest among legacy companies, Li said, “With the help of sensors and algorithms, a self-driving car reduces accident risks, and gives us more time to work, socialize and just relax, while commuting. What happens when 1000s of people have self-driving cars, through the co-ordination of vehicles traffic congestion is reduce, parking is dramatically simplified, as technology reaches more people, its impact becomes more profound”. Google’s ambition in leading AI race is also evident by its significant investment in research, impressive team of researchers who have authored best scientific papers at top AI journals. The results of their work are quickly turned into products and services. A good example of AI investment is the Google Brain program that recognizes expertise in AI and how it will come to be an important resource years ahead.