With Cloud AutoML, Google paves the way for cloud-based AI
Artificial intelligence, which encompasses machine learning, predictive analytics, and deep learning, moved to the forefront in 2017. No longer the stuff of science fiction, companies are using AI technology to gain insight into customer behavior, predict trends, and find ways to be more efficient and save money. One of the big AI trends expected for the very near future is cloud adoption, according to CIO—and Google has poised itself at the forefront. Earlier this year, Google rolled out Cloud AutoML, which uses customer data to generate a machine learning model for that data.
Google is seen as the leader in AI cloud services, according to Geekwire. This comes as no surprise to experts. “They are usually the first to provide a ground-breaking technology, which we have seen in natural language understanding, machine translation, and other fields,” said Konstantin Savenkov, CEO of Intento.
In fact, Google was one of the first to open up its platform for AI, according to a Capgemini report titled “Turning AI into concrete value: The successful implementer’s toolkit.” It open-sourced its TensorFlow platform in 2016 to entice developers onto its platform. Facebook then followed suit with its deep learning framework, Caffe, and Amazon was close behind with MXNet.
While Google has primarily focused on general purpose products, which has led companies to build on other platforms, Google Cloud AutoML has the potential to disrupt the way solution providers build their models, Savenkov said. It’s likely that the new service will enable niche companies to build on the Google stack easily, delivering deeper AI capabilities.
Indeed, one of the big bonuses of Google Cloud AutoML is it drag and drop capabilities, which lets users train their own machine learning clouds. For example, users can line up a set of pictures, then watch as the software picks out recurring elements. However, it’s still in its infancy, and it remains to be seen how capabilities like this will bring Google Cloud closer to competing with Microsoft Azure and AWS.
Yet experts caution that the Google Cloud Platform itself is geared toward programmers, not average users, which may or may not bleed into Google Cloud AutoML as it matures. “GCP is almost built for Google itself and extended to the general public,” said Anna Knezevic, managing director at M&A Solutions.
Aside from that, the back end of the Google Cloud may prove to be the most useful when working with AI technology. Google itself touts its hardware accelerators and analysis, and experts agree that the company has built a platform that can outperform its competitors. Google is definitely best at providing machines and environments that are optimized for heavily quantitative tasks, Knezevic said.
The best example of this is the Google Cloud TPUs, the accelerators and hardware optimized for machine learning, according to Knezevic. Amazon does not have this specific chip that outperforms GPUs and is tailored for heavy-duty matrix calculations, she said.
Whether or not Google is truly the best for AI technology remains to be seen. However, according to experts, Google’s reputation for innovation, combined with its strong back-end processing, may very well mean it will lead the pack in cloud-based AI.
Christine Parizo is a freelance writer specializing in business and technology and has written extensively about SaaS, cloud integration, DevOps, marketing technology, and manufacturing technology.