Google’s new machine learning service targets businesses

  • Machine learning vision service designed to use businesses’ data sets
  • In dead heat with AWS, and others over who gets to dominate AI, machine learning

 

Google’s new machine learning service targets businesses

 

GOOGLE Inc has launched what it claims to be the first of its kind machine learning service which enterprises with little or no knowledge of artificial intelligence (AI) can easily tap into and enable them to start building their own high quality, custom machine learning models.

Globally available from Jan 18 the service, dubbed Google Cloud AutoML, is being touted as a tool to help businesses with limited AI expertise start building machine learning models using advanced techniques such as learning2learn, and transfer learning from Google.

Speaking to journalists on a globally-linked teleconference link, Google’s chief scientist for AI and machine learning Fei-Fei Li (pic, above) said the service is part of the search giant’s main mantra “to organise the world's information and make it universally accessible and useful.”

Li explained that a subset of this overarching goal is to democratise AI and machine learning – to make AI accessible to all – and lower the barrier entry for developers, businesses, and researchers, so that they can make AI useful according to where it fits their respective needs.

Machine learning is a subset of AI, where software is taught to recognise patterns through the observation of massive data sets automatically, often without being explicitly programmed by humans.

Its goal is to automatically allow computers to take a certain action or draw certain conclusions from a large set of data, where the decisions made are automatically refined and evolved given more data analysed.

Examples of these can be found in how Google Photos allows users to upscale their low-resolution pictures to become better images. Or how Google’s anti-spam filters powered by machine learning are able to successfully capture rogue emails.

Arguing that even many large enterprises and companies today do not have the resources, budgets, and expertise to make a leap into the application of AI, Li said Google’s goal is to make machine learning services useful for all.

The introduction of Cloud AutoML is another significant step Google is taking to further make highly-specialised tools available to all, Li said.

“There are a very limited number of people that can create advanced learning models,” she said in a blog post announcing the launch.

“Even if you’re one of the companies that have access to machine learning expertise, you will still have to manage the time-intensive and complicated process of building your own customer machine learning model.”

Li claimed that while Google had previously offered pre-trained machine learning models in the form of its Cloud Vision API (application programming interfaces), there is to date no service that offers the building of custom models based on data sets fed by customers.

Google’s new machine learning service targets businessesExplaining this further, Google’s senior director and head of research at Google Cloud Jia Li (pic, right) said, “Google Cloud Vision API offers a generic vision model that is trained by Google and is offered as an API for customers to use. 

“If customers don’t have any training data, they can use the Cloud Machine Vision API and get direct results from this service using its pre-trained model."

Jia Li said the difference with Cloud AutoML is that Google is providing a specific way in which a model can be trained by using a customer’s data set instead.

“For example, a generic vision model [using Cloud Vision API] that classifies a dog and a cat isn’t the same as classifying a red shoe over a green shoe [in the context] of a retail business.

“By offering Cloud AutoML, we really want to give flexibility to our customers to define what kind of concepts they can classify and what kind of data they can bring to train the model.”

Next page: The differentiating factor

 

 
Keyword(s) :
 
Author Name :
 

By commenting below, you agree to abide by our ground rules.

Subscribe to SNAP
Download Digerati50 2018-2019 PDF

Digerati50 2018-2019

Get and download a digital copy of Digerati50 2018-2019