Google has designed its own computer chip for driving deep neutral networks, an AI technology that is rapidly reinventing the way Internet services operate. This morning, at Google I/O, the centerpiece of the company's year, CEO Sundar Pichai revealed that Google has designed an ASIC, or application-specific integrated circuit, that's specific to deep neural nets. These are the networks of hardware and software that can learn specific tasks by analyzing vast amounts of data. Google uses these neural nets to identify objects and faces in photos, recognize the commands you speak into Android phones, or translate text from one language to another. This technology has even begin to transform the Google search engine.

Google had a lot to showcase at its annual developer conference on Wednesday, from a new home virtual assistant to plans to delve deeper into virtual reality. This year's Google I/O focused heavily on machine learning and VR, and how these technologies are being applied to Google's core products. Google Home, launching later this year, is very much what it sounds like: a virtual assistant for your house. It's a voice-activated device meant to be placed in the home that allows for access to Google when your phone isn't nearby. The company also positioned the device as a smart controller for the home during its presentation, saying it's capable of communicating with nearby connected gadgets like smart light bulbs and Nest devices.

Jefferson Graham reports from Google's I/O developer conference, where the Internet giant introduced new home products and apps aimed at having Google make your life easier. The standalone device will compete directly with Amazon's popular Echo and should be available to consumers later this year, Google CEO Sundar Pichai announced at the search company's annual developers conference, Google I/O. The move had been anticipated, as Google looks to put its mark on a coming age of artificial intelligence in which machines learn to interpret and answer human queries by leveraging the speed and scope of cloud computing. Though smartphones still hold a lock human-computing interactions, the surprise hit of Amazon's Echo speaker has energized a new category: speakers run by digital assistants connected to a tech giant's app and content ecosystem. Google Home project lead Mario Querioz held the device in his palm, revealing a design that was shorter and wider than Amazon's cylindrical Echo, which is powered by Amazon's virtual assistant Alexa.

It can be a pain to install phone apps you know you'll use just once or twice. The app runs on Google's servers instead of your phone. Only the parts you need get sent to your phone on an as-needed basis. If it works as Google envisions, without lags and other annoyances, users won't have to spend a few minutes downloading and installing that app and having it take up valuable space on the phone. The app maker needs to enable this feature, though.

From virtual reality to a new smart-home speaker, Google is showing off just how pervasive it has become even as it's squeezed by its biggest competitors -- Facebook, Apple and Amazon. Google showed off a VR system called Daydream, along with plans for headsets that will compete with Facebook's Oculus Rift. In a jab at Amazon, the company announced Google Home, an Internet-connected speaker that listens for your voice commands to play music or control lights and thermostats in the home. It is reminiscent of Amazon's Echo and will be available later this year for a yet-unannounced price. In an attempt to outshine Apple, Google is also adding features to its Android operating system, including the ability to run apps without actually installing apps.

Forget the CPU, GPU, and FPGA, Google says its Tensor Processing Unit, or TPU, advances machine learning capability by a factor of three generations. "TPUs deliver an order of magnitude higher performance per watt than all commercially available GPUs and FPGA," said Google CEO Sundar Pichai during the company's I/O developer conference on Wednesday. TPUs have been a closely guarded secret of Google, but Pichai said the chips powered the AlphaGo computer that beat Lee Sedol, the world champion in the incredibly complicated game called Go. Pichai didn't go into details of the Tensor Processing Unit but the company did disclose a little more information in a blog posted the same day as Pichai's revelation. "We've been running TPUs inside our data centers for more than a year, and have found them to deliver an order of magnitude better-optimized performance per watt for machine learning. This is roughly equivalent to fast-forwarding technology about seven years into the future (three generations of Moore's Law)," the blog said.

Google is no stranger to building hardware for its data centers, but it's now going so far as to design its own processors. The internet giant has revealed the Tensor Processing Unit, a custom chip built expressly for machine learning. As Google doesn't need high precision for artificial intelligence tasks, the TPU is focused more on raw operations per second than anything else -- it's an "order of magnitude" faster in AI than conventional processors at similar energy levels. It's space-efficient, too, fitting into the hard drive bays in data center racks.

In the end, I suppose it's less interesting to me to look at the sheer amount of machine learning hype than at its content. Like, almost everyone in the 1950s knew that computers were going to be important, and of course they were right, but they were often wildly wrong about the reasons (e.g., dramatically underestimating the difficulty of humanoid robots, while failing to foresee PCs or the Internet). There's no doubt in my mind that people 30 years from now will agree with us about the central importance of ML, but which aspects of ML will they rage at us for ignoring, or laugh at us for obsessing about when we shouldn't have? I don't know the answers to those questions, but I know that those are the things I'd like to know.

In contrast, unsupervised learning systems freely analyze'patterns' in unlabeled data, with no corresponding error or reward linked to a conclusion. It works with'unlabeled data' and is similar to'associative' or'discovery' learning in humans, something that we do very well (and often take for granted). For example, when an unsupervised system is asked to sort or arrange fruits based on raw observations, the system might'choose' to arrange the fruit based on recognition of color, placing strawberries and cherries in the'red' category; or, the system might sort based on observed sizes, grouping pears, apples, and oranges in a'medium-sized' fruit category. This latter method is commonly known as'clustering' and the accepted approach used by these systems to categorize information. Unsupervised learning is a stepping stone, a means to another end such as categorization or finding potential correlations or solutions unable to be spotted by humans or supervised learning systems alone.

The core uses for Wear so far are glanceable information, messaging and fitness. Each of those parts of the OS have been improved, but the changes actually reach far beyond just that. "For the very first time, we've been able to take a holistic pass across the design of the entire system and UI to really hone and tune the interactions around key things that people want to do," Singleton says. Some of the most profound changes to Wear come under messaging, so let's start there. Many of the changes Singleton outlined go far beyond messaging apps, most notably notifications in general.