Interest in machine learning has exploded in recent years as companies realize it has applications in photography, self-driving cars, games, and more. We’ve reached a point that there aren’t enough experienced programmers and data scientists with the necessary expertise to build these systems. Google’s solution is Cloud AutoML, a point-and-click system for building machine learning models without any coding experience.
Google has long offered pre-trained neural networks accessible via APIs that can perform certain tasks, but that’s only useful if you need exactly what that model does. The gist of Cloud AutoML is that almost anyone can bring a catalog of images, import tags for the images, and create a functional machine learning model based on that. Google does all the heavy lifting behind the scenes, so the customer doesn’t need to know anything about the intricacies of neural network design.
AutoML won’t compete with the cutting edge, highly tuned AI systems an experienced engineer could build, but few businesses have the money or resources to support the development of completely custom machine learning models. AutoML uses a simple graphical interface, allowing the user to drag in a set of images. Then, the platform needs to know how to describe those images. Google does its magic, and you end up with a model running in the cloud that can identify the specified terms in photos. AutoML provides stats on the strength of the model, so you can train it with more data or test with new images.
The end result is a machine learning model that runs on Google’s servers, accessible via an API. Users can reach out to that model via the Google cloud API and get predictions on new images. For example, both Disney and Urban Outfitters have tested AutoML to identify objects in their online stores so users can search and filter with more terms. Thus, you could search for “blue backpack” on Urban Outfitters and see all the blue backpacks, even if the items were not tagged that way in the system.
Google’s Cloud AutoML is currently limited to images, and it’s is alpha. You need to apply for access to the alpha version, and there’s no guarantee you’ll get in right now. The vision part of AutoML is just the first part of several features planned for the product. Google did not mention cost, but it’s likely businesses will have to pay for API access to the models they create in AutoML.
Google’s Cloud TPU Matches Volta in Machine Learning at Much Lower Prices
Google and Nvidia both offer competitive machine learning products — but Google is beating Nvidia on costs, at least in certain tests.
Intel Positions Xeon as Machine Learning Competitor in Inference Workloads
Intel is tossing its own hat into the machine learning ring, positioning Xeon as a potent inference workhorse.
Google Announces 8x Faster TPU 3.0 For AI, Machine Learning
Google's new TPUs are here — and they're quite a bit faster than last year's model.
Machine Learning Works Wonders On Low-Light Images
New work from Intel and the University of Illinois Urbana–Champaign demonstrates some incredible advances in low-light photography, courtesy of machine intelligence.