CRN Exclusive: How Google Is Using AI And Why Partners Are 'Absolutely Crucial' To Its Strategy

Standing Out From The Pack

Google Cloud Platform is one of the industry-leading providers in the hyperscale public cloud computing market, but that doesn’t mean that the cloud heavyweight can rest on its laurels. Faced with stiff competition from cloud giant Amazon Web Services and Microsoft Azure, Google has had to prove to business customers that it offers more than cheap storage, a popular email platform, and some niche collaboration tools. Artificial intelligence (AI) and machine learning, are helping the provider stand out.

Rajen Sheth, senior director of product management for Google and leader of the company's cloud AI and machine learning team, filled CRN in on how machine learning and AI are helping Google land more enterprise customers and why channel partners are critical to the cloud giant's AI strategy.

Here are excerpts of CRN's conversation with Sheth.

How is AI and machine learning helping Google differentiate from its cloud competitors?

[AI and machine learning] really are among the key differentiators for us. We talk a lot in the media about Google as an AI-first company, and it's much more than just a marketing message. What has happened is all of Google has really shifted over the last three years to be really AI-centric. That includes completely revamping the infrastructure we have to be ready for AI, and building amazing models that do incredible things with AI and integrating those into applications. As Google Cloud, we get to take advantage of all of that.

In our data center, we have custom-designed silicon specially for machine learning that runs much faster than anything that's out there now. We released the TPUs [Tensor Processing Units, an internally designed chip series specifically built to power artificial intelligence workloads] in beta to cloud customers [in February] and we are seeing really amazing results.

Can you tell us about TensorFlow?

TensorFlow [a computational framework for building machine learning models] is kind of the secret sauce of the machine learning platform that we have internally, and we've open sourced that so everybody has access to it. That's something we are continually innovating on, and we have it running on Google Cloud. We also have best in class facilities for things like image recognition, video, speech to text, and natural language processing. We use those in a lot of products for sure, but now we provide all of those "as a service," so any business customer can come and actually start to use those services as part of their application.

How are you helping users new to AI and machine learning get started?

One thing I think that really distinguishes us is AutoML [Google's portfolio of machine learning products]. A lot of businesses came to us with their problems they were trying to solve with machine learning, and they were trying to do something a little bit customized, so what we have now to help with that is AutoML; a [service] that creates machine learning models. We released AutoML in January, and it's really democratized AI and is bringing it to many more people.

What are the biggest obstacles for businesses in implementing AI and machine learning?

One problem we are running into is companies don’t have the skillset to build machine learning models. Things like AutoML can really make it so the average developer who doesn’t have that in-depth machine learning knowledge can start to use machine learning in very tangible ways. We are seeing people come to us because of our machine learning capabilities, but then they start to move rest of their workloads as well, so it's becoming a key way to attract users to Google Cloud.

Where do channel partners fit in to Google's AI and machine learning strategy?

Partners are absolutely crucial to our AI and machine learning strategy - maybe more so to this strategy than any of our other cloud technologies. The partners have domain expertise that we don’t have. Some have really good machine learning skills and they can go to the platform and start building their own models on top. But a lot of them don’t, and this opens up opportunity for partners to do some interesting things with machine learning for their customers. For example, With AutoML, a lot of [solution providers] are using to it create custom solutions for their customers and solve particular business problems. In my mind, the only way we are going to scale is through partners.

The way machine learning differs from cloud infrastructure is with cloud infrastructure, you can take a virtual machine from one place and move it to another place, so it’s a fairly generic thing. With machine learning, not only do you have to implement the model, but you have to make sure the data is in the right place and you have to change the workflow of people using it .... really only partners have ability to do that and that’s where we see a lot of opportunity for partners to be able to bring this to market.