Google races against AWS, Microsoft to bring AI to developers

AI has always been central to Google Cloud's value proposition and is a major theme at the Google Next conference, but AWS and Microsoft have their own plans for making AI more accessible.
Written by Stephanie Condon, Senior Writer

The Cloud in 2018: what we have learned so far

At the Google Cloud Next conference in San Francisco Tuesday, Google laid out how it's bringing artificial intelligence to developers, as well as integrating more AI capabilities throughout its cloud products.

Artificial intelligence has long been a cornerstone of Google Cloud's value proposition, but to win more customers it needs to make those capabilities more accessible. It also has to contend with Amazon Web Services and the fast-growing Microsoft Azure, which have been building up their own AI-powered offerings and creating their own plans for lowering the barrier to entry.

During the Day One Next keynote, Google Cloud CEO Diane Greene noted that Google is heavily investing in two key areas: AI and security. While Google is investing in security because it is customers' "number one worry," it's investing in AI because it is the "number one opportunity."

AI is "key to re-engineering a business," Greene said. "Today it's built into everything Google does. We are now working to make it easy for you. We are incorporating AI into everything you do."

To make AI more accessible, Google announced the expansion of Cloud AutoML, the software that automates the creation of machine learning models. Announced earlier this year, AutoML makes it possible to build custom machine learning models without any specialized machine learning knowledge. It effectively extends Google's Cloud Vision API to recognize entirely new, customized categories of images.

Google in January announced the alpha of AutoML Vision, and on Tuesday it announced the product is moving into public beta. This means any Google Cloud customer can submit a set of labeled images, and Google will create an image recognition model matching that data set. Since announcing the product in January, around 18,000 customers have expressed interest in AutoML Vision, Rajen Sheth, senior director of product management for Google Cloud AI, told reporters.

Additionally, Google is introducing AutoML Natural Language and AutoML Translation. With AutoML Translation, customers can build models that take into account industry-specific language. For example, the phrase "the driver is not working" would be translated differently for the computer industry than it would be for the transportation industry.

In addition to expanding AutoML, Google on Tuesday launched updates to its machine learning APIs. The Cloud Vision API now recognizes handwriting, supports PDF and TIFF files, and can identify where an object is located within an image.

The Cloud Text-to-Speech API is getting updates that include the ability to optimize for different speakers from which the speech will play. Meanwhile, Cloud Speech-to-Text can now identify what language is spoken as well as different speakers in a conversation. Multi-channel recognition enables users to record each participant separately in multi-participant recordings.

Google's AI hardware is also getting an update: The third generation of Google Cloud TPUs are now available in alpha. Second generation TPUs are now generally available, meaning all GCP users can access them, including free tier users. TPUs, Google says, dramatically accelerate machine learning tasks and are accessible via GCP.

In an indicator of how AI may be more ubiquitous in the future, Greene on Tuesday said that Google sees a disproportionate amount of TPU use from startups on the Google Cloud Platform.

Google also announced updates to G Suite that included a heavy dose of artificial intelligence.

Google's competitors are also offering ways to put machine learning and AI in the hands of developers. Late last year, AWS unveiled SageMaker, which makes it easier and faster to train machine learning models. At the AWS Summit in New York earlier this month, the company announced it's bringing streaming algorithms as well as batch job improvements to the service,. AWS also used the summit to bring DeepLens, a deep learning enabled video camera for developers, into general availability. The general takeaway of the summit was that Amazon Web Serivces is a platform designed for analytics, AI and machine learning.

Meanwhile, this year's Microsoft Build conference, the company's annual developer confab, was filled with data- and AI-related announcements and demonstrations.

Editorial standards