Every time the big three cloud vendors — Google, Microsoft and Amazon — hold an event, we look for the onslaught of new tools aimed at developers, IT providers, and data scientists. It seems like with Amazon so far in the lead, that it may be all window dressing. But it definitely matters and each company is trying hard to make its case.
Google had its I/O developer event last week and had so many announcements, it was hard to keep up without a scorecard, but one thing is clear. Google wants to be the artificial intelligence/machine learning cloud. It started with an update to its custom Tensor Processing Unit chips, which are supposed to faster than the fastest CPUs and GPUs on the market.
How fast? Well, according to CEO Sundar Pichai each board is capable of 180 trillion floating point operations per second. You don't have to be a computer scientist to get that, that's a lot of operations. But they didn't stop there.
They designed them to put into groups of 64 boards, which creates a supercomputer in the cloud they call TPU pods. Each pod is capable of 11.5 petaflops, an astonishing amount of computing power packaged as a cloud service, putting it within reach of just about anyone who has the means to pay for it (and certainly a lot closer than having to buy, install, and maintain a comparable system if it were available on the open market).
Google was also announcing intelligence built into a variety of tool including Google Lens, a tool like Google Goggles that can identify what you are looking at by pointing your camera it and Google Photo, which can now among other things recommend people to share photos with based on who is in the picture. There were updates to Google's voice assistant, Google Home that allows it to push notifications to you. In the example, Google provided, Google Home would signal when it has information such as suggesting you leave earlier based on a traffic issue it could infer from reading your calendar and where you needed to be (which is pretty amazing).
While that might be more information than some of us want to share with Google, the fact is that Google and its competitors are building artificial intelligence into just about every product. What's more, the cloud is a key part of all of this.
Google doesn't just want to make it easier to use its vast array of tools by souping them up with AI — although that's clearly a big part of the strategy — they also want to expose vast computing power as a cloud service. While they didn't talk a lot about their machine learning APIs at this event, just as the company created products like Kubernetes, TensorFlow, and Spanner from in-house projects, it follows that at some point Google will package at least some of this underlying artificial intelligence it's building into its products as a service too.
Google appears to aspire to be the machine learning and AI company, although its competitors might have a thing or two to say about that. Regardless, it still needs to show it can pull all of the pieces together into a coherent set of products, while competing hard with AWS and Microsoft, who don't forget, are continually making counter moves of their own.
Photo: Bruce Washburn on Flickr. Used under CC by 2.0 license.