Google mixes machine learning technology with cloud platform

Cloud Machine Learning gives companies easy access to smart technology

SAN FRANCISCO: Google is putting its machine learning technology and services into the cloud with the unveiling of Cloud Machine Learning (CML).

Jeff Dean, a senior fellow at Google Research, said at the Google Cloud Platform (GCP) Next event in San Francisco that the suite of machine learning tools is built on top of the firm's TensorFlow open source intelligent software library and offers the ability to tap into cloud-power translation, image recognition and data analysis services.

CML is effectively a packaged suite of machine learning APIs that Google already uses in the Photos and Translate services, as well as voice recognition developed for Android, delivered as services through GCP.

The tool is aimed at application makers looking to embed machine learning into apps and companies looking to apply machine learning analysis to datasets. It provides pre-built machine learning models and allows users to build their own versions for specific uses.

"It's really important to help bring the power of machine learning to all kinds of different users internally and externally," said Dean.

"This product will allow you to understand your data, visualise it and prepare it for machine learning using the other tools that already exist on our cloud. And then you can use the power of our system to solve problems in your particular context, and take those models that you've developed and deploy them at scale."

Delivering the APIs through the GCP gives developers a way to tap into machine learning capabilities without needing the compute power and infrastructure to support the data crunching that underpins machine learning.

This simplifies access to complex technology normally reserved for major companies and research organisations.

Google's main rivals in the cloud platform arena, Microsoft Azure and Amazon Web Services, already offer machine learning services with Azure ML and Amazon Machine Learning.

But a panel of Google executives told V3 that Google could stand out from its rivals with a machine learning technology that has beaten a champion Go player.

CML has also been developed in a way that allows it to scale with demand through Google's extensive cloud infrastructure. Furthermore, the company will allow machine learning models developed with the services to be exported for use on other systems and on-premise infrastructure.

Google's push will further encourage the rise of machine learning technology, which is increasingly being offered as an easy to access cloud service.