Google I/O 2017: AI, IoT and VR/AR predictions

Artificial intelligence will be a major theme as Google focuses on growing its new business segments: cloud, IoT, VR and AR

Google I/O 2017: AI, IoT and VR/AR predictions
Google

Google I/O 2017, Google’s other annual developer conference, begins next week (May 17). It is the other developer conference because Google filled the Moscone Center with 10,000 enterprise cloud developers at its Cloud Next conference last March. Compared that to the 7,000 attendees at Google I/O 2016. The two conferences explain two different developer audiences and Google’s cloud growth ambitions.

The list of code labs at Google I/O 2017 confirms this: accessibility, ads, Android, Android devices, Google Assistant, Firebase, IoT, location & maps, machine learning & AI, Flutter, mobile web, Google Play, virtual reality. Though many developers attending I/O will attend both conferences, this is a much different schedule than Cloud Next.

+ Also on Network World: Star Wars Holochess can become a reality thanks to Vuforia and Tango +

Google’s cloud platform will not be absent from Google I/O, however, and there could be a very big cloud machine learning announcement. Also, Google’s Firebase cloud development platform will get a lot of attention. Firebase fits I/O’s audience because it is a unified mobile and web development platform that abstracts much of the backend cloud integration for app developers. Firebase delivers development efficiency and fast deployment that addresses a large well-defined problem, real-time data synched across multiple mobile and web platforms. 

There has not been much speculation or leaks about new product announcements for this year’s I/O. But since Google announces new products at every I/O, the company can be expected to debut more than just dot releases of existing products.

Looking over the schedule, the event will be very mobile-, Android- and IoT-focused. Most of the web development will be delivered as web and mobile cross-platform development platforms and tools. The developer preview release of Android O was introduced a month ago so developers could prepare for I/O. But next week, we will learn what O stands for. If Google continues the theme of Android releases being named after sweet things, it might be Oreo.

Speculation of I/O product announcements

Google goal is to expand into new markets, AI and machine learning, including assistant devices and assistant capabilities on Android and iOS devices and bots, virtual reality (VR), augmented reality (AR) and IoT. These are the possible areas in which Google will announce new products.

New AR and VR products, an AR headset and 6DoF VR

Google Tango’s camera and sensor assembly and Android could be built into an AR headset. The Tango assembly is in its third release and is the most stable 3D-sensing system of all the AR devices. The only thing it is missing is the ability to project animated holographic objects that interact with the planes and objects in the Tango headset user’s field of view.

Two weeks ago, Vuforia announced support for Smart Terrain that delivers this capability on Tango with Google and leading AR/VR authoring engine maker Unity3D. All that is missing is reconfiguring Tango into a headset form factor.

Google Daydream VR currently supports 3 degrees of freedom (3DoF). Simply explained, 3DoF means that when someone wearing the Daydream VR head-mounted display walks, the VR experience does not change to reflect their change in position.

With 6DoF, when a user changes position, the field of view changes. For example, a user in 360-degree Street View could walk down a virtual street like they could in reality. Using a derivative of the Tango assembly and software, the user’s virtual position in a Daydream app could be mapped into real 3D space, and if the user is about to bump into a chair or wall, the user could be warned.

Allo chat app will be opened to developers with Google Assistant 

Personal assistants and bots are a growing market. The Google Home personal assistant has had a year to mature and has caught up with Amazon’s Echo, giving Google plenty of experience building bots internally. Facebook and Microsoft have released bot building messenger tools, too.

Allo and Google Assistant development tools could be merged for developers to build intelligent social apps to augment conversations and for commercial apps. Right now, Allo predicts text and sticker replies, makes recommendations such as restaurants, and interfaces with Google Assistant to access the company’s semantic knowledge graph of over a half billion objects. Google could announce that Google Assistant will run on more Android devices and that developers will get access to it through APIs to build conversation actions for single-user experiences and through an Allo API to build multi-user experiences.

Google Home, more IoT devices

Google will announce a lot more IoT devices that are controlled from Google Home. Developer activity with Android’s IoT platform, Brillo and Weave has been high. But each device interfaced to Home is a collaborative effort by Google and the device maker to protect users and Google’s brand. This is limiting because Google will only work with as many device makers as its OEM staff can handle at one time.

Google already protects Google Home with strict brand guidelines and by requiring their approval of Home apps called Actions. Once Google approves an Action, the app is available to everyone with a Home device without the need to download and install it.

Google may open more access to Home by publishing an interface to Home to allow IoT device makers to build and test new products without the ability to distribute them until they are approved as part of the Actions approval process. This would stimulate the IoT device makers to create more devices.  

More Tensorflow models that run on Android, Android Wear and IoT devices

At the Tensorflow Summit last February, Google demonstrated Tensorflow models running on Android. It is a non-trivial task to train models on a more powerful machine that runs on small memory devices to add AI to the device. Google could announce tools that developers can use to reduce the size of models to run on small memory devices.

A new machine learning model and new training and professional services products

This is probably the most speculative prediction. But here is the explanation. Machine learning, in a way, has been defined by the interests of academic researchers. And big cloud company research investments from Amazon, Google, Facebook, Microsoft and IBM reflect the very large-scale web features and automation machine learning problem definition. Most development teams outside of this small group of companies do not have machine learning experts to build models. Instead, they use prebuilt models such as natural language recognition, language translation, image recognition and ranking that were the products of academic research and developed by these large companies.

Growth will come from expanding the number of developers building with pre-built models and delivering new models that solve a newly defined machine learning problem. Google has had a lot of experience distributing machine learning from Google Brain to its product developers.

The first and most likely prediction is that Google will announce training courses, professional services, and tools that abstract data scientist types of tasks, such as normalizing data for training machine learning. Googler Christine Robson’s applied machine learning team has gained much experience distributing R&D to Google product groups that could be applied to new courses and professional services.

The second and less likely prediction is a new machine learning model. Late last year, machine learning research luminary and head of Stanford’s AI research, Fei-Fei Li, Ph.D., joined Google’s Cloud Group Platform group. One of Li’s focuses is to find and build new models and apply machine learning to new and large green fields. Whether this new model will apply to a specific industry or across industries is hard to predict. If this prediction does not come to fruition, it will not be wrong, only early.

Google’s software portfolio has become so large it really has two developer conferences Cloud Next and Google I/O. And if the Tensorflow summit is produced again next year, that will give it three. AI will be a major theme in both as Google focuses on growing its new business segments: cloud, IoT, VR and AR. 

Join the Network World communities on Facebook and LinkedIn to comment on topics that are top of mind.

Copyright © 2017 IDG Communications, Inc.