The company's CEO, Sundar Pichai, spoke about Google's transition to the era of Artificial Intelligence, opening the Google developer conference, "I/O 2017".
In his speech, he said: "I've been at Google for 13 years, and it's remarkable how its founding mission, to make information accessible and useful, remains as relevant today as ever. Since the beginning, we've applied computer science and its insights to solving complex problems, even as the technology around us has forced dramatic changes.".
The most complex problems tend to be the ones that affect people’s everyday lives, and it’s fascinating to see how many people have made Google a part of their daily lives: There are 2 billion active Android devices, YouTube has not only a billion users but also a billion hours of watch time every day, and Google Maps is used to navigate 1 billion kilometers every day. This growth would have been impossible without the shift to mobile computing, which has led Google to redesign all of its products—reinventing them to reflect new interaction models, like multi-touch screens.
We are now witnessing a new shift in computing: the shift from mobile dominance to artificial intelligence. As in the past, this shift is forcing Google to reinvent its products in an effort to create a world that allows for more natural and seamless interaction with technology. Think of Google Search: it was built on the ability to recognize text on web pages. But now, through the application of machine learning, we can make images, photos, and videos more useful to users than ever before. Your phone’s camera can “see,” you can talk to your phone and it will respond—speech and vision are becoming as important to computing as keyboards and multi-touch screens.
Google’s digital assistant is a tangible example of these developments. It’s already available on 100 million devices, and it’s getting more useful every day. Google Home can now distinguish between different voices, giving users a personalized experience every time they interact with the device. Google is also now able to turn your phone’s camera into a tool for performing everyday tasks.
Google Lens is a set of computing capabilities that , by understanding what you see, can help you take your next actions. For example, if you're on the floor of a friend's apartment to see a long and complicated Wi-Fi password behind a router, your mobile phone can now recognize the password, understand that you're trying to connect to a Wi-Fi network and automatically connect you. The key point is that you don't have to learn anything to do it. The interaction and experience becomes much simpler than the process of copying and pasting passwords from one application to another on a smartphone. The new Google Lens capabilities will be added to the Digital Assistant and Google Photos first , and then you can expect them to be available in other products.
All of this requires the right computing infrastructure, however. Last year at I/O, Google announced the creation of the first generation of dedicated TPU (Tensor Processing Unit), which allow machine learning algorithms to run faster and more efficiently. Today, we are announcing the next generation of TPU processors: Cloud TPUs, which optimize inference operations, the machine learning process, and can process much more information. Google is bringing Cloud TPUs to its platform and Google Compute Engine, so that companies and developers can benefit from it.
It's important for us to make all of these developments beneficial to everyone, not just Google users. Great breakthroughs in complex social problems can become a reality if scientists and engineers have better, more powerful information tools for their research. But today, there are too many obstacles to making all of this a reality.
That’s the motivation behind Google AI, which is merging all of Google’s AI initiatives into one that can break down these barriers and accelerate the efforts of researchers, developers, and companies.
One way we hope to make AI more accessible is by simplifying the creation of machine learning models, also called neural networks. Today, designing neural networks is very time-consuming and requires such cognitive expertise that it limits their use to a small community of computer scientists and engineers. To that end, Google has created an application called AutoML that demonstrates that it is possible for neural networks to design new neural networks. At Google, we hope that AutoML will gain the ability to design neural networks that only PhD holders in the relevant field currently have, and in 3 to 5 years, offer the ability to hundreds of thousands of programmers to design neural networks based on their needs.
In addition, Google AI researchers (Google.ai) collaborate with scientists from many academic disciplines and developers to find solutions to a range of problems, with very satisfactory results to date. We used machine learning to improve the algorithm that detects the spread of breast cancer to nearby lymph nodes. We also found that artificial intelligence minimizes the time and maximizes the accuracy with which researchers can explore the properties of molecules and human genomes.
This transition to the new world of Artificial Intelligence is not limited to simply creating futuristic devices or conducting cutting-edge research. We believe that it can help millions of people today through the complete democratization of access to information and the emergence of new opportunities.
For example, nearly half of employers in the United States say they have trouble filling job openings. At the same time, unemployed people often don’t know about a job opening right next to them because the nature of the ads—frequent changes, low traffic, inaccurate job titles—has made them difficult for search engines to spot. Through a new initiative, Google for Jobs, we hope to connect companies with potential employees and help job seekers find new opportunities. As part of this effort, we’ll be launching a new feature on Search that will help people search for jobs based on their experience and salary levels. The jobs will include sectors that have traditionally been more difficult to find and rank, such as services and retail.
It is inspiring for us to envision the fruits that AI can yield. There is still a long way to go before we truly reach the world of AI, but the more we can work to provide greater access to technology – by giving people tools – the sooner everyone will benefit.”.
