Tech giant Google has made its "next big leap" in machine learning, launching a new product which lets businesses with little artificial intelligence expertise build and customise their own models using Google's techniques of transfer learning and learning2learn.
Google's first iteration of its new Cloud AutoML product will focus on image recognition and will be rolled out gradually over the coming weeks to customers globally, starting from Thursday.
Known as Cloud AutoML Vision, the product will let users utilise a drag-and-drop interface to upload images, train and manage models and then deploy those machine learning models on Google Cloud.
It has already been trialled by Disney, Urban Outfitters and the Zoological Society of London and throughout 2018 the company also intends to expand the service to other areas such as speech, translation and natural language processing.
Google senior director of product management for Google Cloud artificial intelligence, Rajen Sheth, said Cloud AutoML Vision would be able to offer a higher level of accuracy and faster production time for machine learning models than any other product like this on the market.
"We think this will be the next big leap for us in machine learning," he said.
"It's a very, very simple experience and you end up with a high-quality model. We've started with vision because one of the things we've found if you can get to high accuracy, the use case for vision is numerous."
Increasingly important
With this innovation Mr Sheth identified industries such as media, healthcare and retail as obvious use cases.
Within minutes businesses will be able to have a demo version of an industry-specific machine learning model up and running and within a day a full-scale, production-ready model can be ready to be rolled out.
Artificial intelligence has been emerging as an increasingly important area for Google for the last few years, with the company making its TensorFlow software used for building AI tools open source in 2015. It also offers off-the-shelf products for speech transcription and image recognition, but these algorithms are often limited in their scope.
For example, it could identify that a car is a car, but not the make and model.
The Cloud AutoML product has been designed to address this and will allow businesses and people with specific use cases to easily create their own machine learning algorithms.
This is part of a goal set by Google head of research and development for Cloud AI Jia Li and Cloud AI chief scientist Fei-Fei Li to democratise AI.
New fields
"Currently, only a handful of businesses in the world have access to the talent and budgets needed to fully appreciate the advancements of ML and AI," they said in a blog post.
"We believe Cloud AutoML will make AI experts even more productive, advance new fields in AI, and help less-skilled engineers build powerful AI systems they previously only dreamed of."
Zoological Society of London conservation technology lead Sophie Maxwell said it was planning to use the technology to automate the tagging of photos of animals in the wild.
"ZSL has deployed a series of camera traps in the wild that take pictures of passing animals when triggered by heat or motion. The millions of images captured by these devices are then manually analysed and annotated and with the relevant species ... which is a labour-intensive and expensive process," she said.
"ZSL aims to use AI to automate the tagging of these images – cutting costs, enabling wider-scale deployments, and gaining a deeper understanding of how to conserve the world's wildlife effectively."
http://www.afr.com/yolanda-redrup-j7gdq