The recent Google I/O conference saw the internet giant unveil is artificial intelligence intentions. From modest machine learning beginnings, Google has unveiled products that intend to revolutionise our daily lives and bring machine learning into all current applications.
As the development of mobile applications hit the tech world by storm a few years ago, now is the time of AI and Google intends to capitalise with a regime change. In addition to improving the functionality of Google Home, Search and Photos, the media giant unveiled a new innovation called Google Lens. Home can recognise voices, Search recognises and recommends search results, and what Google Lens brings to the table, is interpreting the surroundings and taking actions based on that information. Along with these innovations, improvements to existing technology have made inroads towards a completely AI focused brand.
Into the future with machine learning
Google may not be the leader of cloud computing, but novel developments have spurred Sundar Pichai, the company’s CEO, to openly recognise this as an economic opportunity and possibly a challenge. The release of second generation cloud tensor processing units (TPU) is making cloud computing and machine learning faster and more efficient.
Google’s TPUs deliver an open source machine learning framework allowing massive scaling and dissemination of applications through the cloud. This innovation will expedite the process of intake, normalization, model training and deployment by essentially using machine learning to automate algorithms. In effect, this is automating deep learning by using machine learning. Google is calling this creation AutoML, and it is the breakthrough of the company’s AI research group, Google Brain.
The new generation of TPUs are designed as unified end-to-end processors that contain the basic framework for machine learning applications. This machine learning “pipeline” will allow developers to build on a standardised machine learning framework, Google’s TensorFlow. However, there has been some trepidation, as although it is a good framework, there are others out there suited to different needs. Thus, some might find the direction of this particular advancement quite limiting.
Google’s machine learning applications
With a reported two billion active Android users, the market is ready and available to exploit these improvements. Successful machine learning relies on large quantities of information. It has been the basis for Google’s search for some time, but has really come to the fore with the launch of Google Assistant last year. Building on the groundwork of the “OK Google” service, this product can not only help you with day to day tasks, but learn your preferences and proffer information based on them.
Google Home was also introduced at the annual developer’s conference. This wireless speaker and smart appliance is set to be released later in the year. It is capable of streaming audio and video as well as controlling smart appliances; turning it into a personal IoT hub. It is compatible with smartphones and thus you will be able to tell Google Home to change your dinner reservations or meetings with lawyers, and the device will amend your schedule accordingly. The demo video shows a father waking his oversleeping son by telling Google Home to turn on the lights in his son’s room.
Smart reply, dubbed “Allo,” has seen some improvements to its suggestions already. Used in Gmail’s “inbox” application, the app can even send a smart reply in response to a photograph using Google’s image recognition software.
Google Photos uses machine learning to recognise people, places and objects, enabling image enhancement and suggestion captions or responses.
However, all of the hype at the conference was building towards the announcement of Google Lens. The vision based computing algorithm is “taught” to recognise surroundings and make inferences and relevant suggestions. For example, a photograph of a restaurant can bring up reviews, ratings and even the menu. In addition, the technology allows automatic connection to Wi-Fi routers. All that is needed is to point the smartphone’s camera to the router’s username and password.
From text-based early developments to the ability to learn and understand images and videos demonstrates how far Google has come. And it’s not stopping now. The company seem to be on a roll with the deployment of new ground-breaking technologies at an increasing rate. Pichai is making no secret of the fact Google is an AI first company and that is where he is hedging his bets.
The machine learning pipeline, enabling swifter development, bodes well for Google’s future in the AI sector, establishing the company at the vanguard of the machine learning revolution.
This article is published as part of the IDG Contributor Network. Want to Join?