Why Google’s Cloud Auto ML is its Most Important ML Launch Yet

Why Google’s Cloud Auto ML is its Most Important ML Launch Yet

While always pushing the forefront of advanced machine learning techniques, Google Cloud’s parallel mission and central problem over the last few years has been: how do they make ML more accessible to users? Google has continuously made significant investments in machine learning and artificial intelligence and it’s the latest product, Cloud Auto ML, the company has made its biggest stride in achieving that goal. Cloud Auto ML is a tool that allows users to build custom machine learning models using your own data and labels without writing a single line of code.

As described by Fei-Fei Li, Google’s Chief Scientist for their Cloud AI and Machine Learning groups, “Cloud AutoML helps businesses with limited ML expertise start building their own high-quality custom models by using advanced techniques like learning2learn and transfer learning from Google. 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.”

After hearing Li’s comments and testing the product ourselves, we see Auto ML helping businesses in three key areas of Machine Learning: ease of use, better models, and speed of model deployment.

Easy of Use (No Code)

Google Cloud Auto ML is more like a SaaS application than a typical ML tool. Creating models is all done through a GUI (Graphical User Interface), so creating state of the art models can be done without writing any code. Through the GUI you can upload data, label data, build a model, view key model metrics, and finally give you a restful endpoint to call your newly created model.

Better Models

Google Cloud Auto ML Leverages Google’s state of the art AutoML and Transfer Learning technology to produce high-quality models. If you’re interested in learning more about what transfer learning is, here’s a great post by Google on the topic. To summarize, AutoML takes out the need for hyperparameter tuning, data augmentation, and a host of other ML related challenges. All you need to do is have labeled data, upload that data to AutoML, and work with the easy to use GUI from there.

Faster Turnaround Time to Production-Ready Models

One of the challenges with applying Machine Learning to business challenges is all the different skill sets required to get a model into a production state. Typically the process requires a data scientist to build the model then the model needs to be handed off to a software developer who will then wrap the model in some sort of secure endpoint. This process is all done for you with AutoML, as a result, speed to production will decrease

Who’s Using It?

While Cloud Auto ML is a fairly new product, there have been a few early adopters who have found success with the tool. Two large companies that stand out are Urban Outfitters and Disney.

Urban Outfitters

“Creating and maintaining a comprehensive set of product attributes is critical to providing our customers with relevant product recommendations, accurate search results, and helpful product filters; however, manually creating product attributes is arduous and time-consuming. To address this, our team has been evaluating Cloud AutoML to automate the product attribution process by recognizing nuanced product characteristics like patterns and neckline styles. Cloud AutoML has great promise to help our customers with better discovery, recommendation and search experiences.”

-Alan Rosenwinkel, Data Scientist at URBN.

Disney

“Cloud AutoML’s technology is helping us build vision models to annotate our products with Disney characters, product categories, and colors. These annotations are being integrated into our search engine to enhance the impact on Guest experience through more relevant search results, expedited discovery and product recommendations on shopDisney.”

-Mike White, CTO and SVP, for Disney Consumer Products and Interactive Media

For further context, Google’s first solution under Cloud AutoML was Auto ML Vision. At Google Cloud Next ‘18 they pushed Auto ML Natural Language and Auto ML Translation into beta and officially added them to the list of solutions underneath the Cloud Auto ML tree. It will be exciting to see how companies use these two new solutions.

Considerations

Cloud Auto ML as a tool today can solve a host of challenges for many companies, however, there are still some considerations to be aware of. The first being the challenge of data quality. Like any machine learning problem, poor data quality or biased data can lead to poor results. Relying on a model without accounting for biases can lead to costly results. As an example, a recent ProPublica report found that a computer program widely used to predict whether a criminal will re-offend discriminated against people of color based on biased training data. With a tool as easy to use as Auto ML, there will be danger in companies relying on these models too quickly. Proper data processes and testing still need to be thoroughly applied with a model produced by Auto ML.

The second consideration to highlight is that while Auto ML gives you a host of new capabilities, it still has some limitations in the flexibility of problems it can solve. For image recognition problems the ability to automatically identify where objects are in an image, and the distance between certain objects, among other challenges, are still problems that will need to be solved by a custom ML model. The two charts below highlight Google’s different offerings in the ML space and the pro’s and con’s of each ML tier.

Making ML Available to Everyone

The demand for machine learning expertise is far higher than the supply in the current job market. With Cloud Auto ML, companies who may not have advanced in-house ML experience can now begin to test and deploy ML models to their business challenges. Auto ML will increase ML adoption across nearly every industry because of its ease of use, increased accuracy, and speed of deployment. As the barrier to entry gets lower for ML and AI, organizations must continue to keep ML at the forefront of their technology roadmaps if they hope to keep pace.

2018-08-27T10:12:39+00:00August 5, 2018|Categories: Cloud & Machine Learning, White Papers|

About the Author:

Matt Powers
Matt Powers has spent the last 2 years in a technology delivery role for Maven Wave focusing on machine learning and big data. Matt recently worked at Google as a customer engineer helping organizations take advantage of Google Cloud Platform's computing, big data, and machine learning capabilities.