Top 10 Things Unfolded in Google I/O 2017

Introduction

Google I/O 2017 marked the conference’s 10th anniversary, and the announcements that week sent a clear signal to every software development team: Google was no longer just a mobile company. With over 100 updates announced across five days in Mountain View, the event confirmed that AI would drive every major platform decision going forward. From San Diego to San Francisco, development teams immediately began evaluating which of these announcements required architectural changes to their existing roadmaps.

What made I/O 2017 different from prior years was not any single feature announcement, but the coherent thread connecting all of them. Android O, Kotlin, Google Lens, the new TPU, and Android Go were not isolated product updates. They were pieces of a unified AI-first strategy that would reshape how mobile and cloud applications are built. Here are the 10 announcements that mattered most for developers and product teams.

Android O: What Did Google Announce for the Next OS?

Android O was the headline announcement, launching in beta at a moment when Google confirmed 2 billion active Android users worldwide. The new OS introduced Picture in Picture, which lets users keep a video or navigation window active while working in another app. This feature opened a new layer of multitasking that developers needed to design for immediately.

Notification Dots gave apps a visual indicator of unread activity directly on the home screen icon, reducing friction for users who wanted quick context without opening an app. Autofill was formalized as a system-level API, enabling credential managers and form tools to hook into Android natively rather than relying on accessibility workarounds. For custom software development teams, Android O required a targeted migration effort to take advantage of these new system behaviors.

Android Go: How Google Targeted the Next Billion Users

Android Go was designed for smartphones running on less than 1GB of RAM, a segment that represented more than one quarter of total Android shipments in India during Q1 2017, according to industry tracking data at the time. Google planned to release the OS the following year, specifically to bring connected software access to lower-income markets across Southeast Asia, Africa, and Latin America.

For development teams thinking about market reach, Android Go changed the conversation around “Lite” app architecture. Smaller APK sizes, reduced memory footprints, and offline-first data patterns became competitive requirements rather than optional optimizations. Teams building mobile app development products for global markets had to rethink assumptions around device capability from that point forward.

Google Assistant: Expanding to iOS and Adding Typing Support

Google Assistant’s expansion to iOS was a direct acknowledgment that the assistant experience could not be limited to Android if it was to serve as a true AI-first platform layer. The iOS app brought voice and contextual AI to iPhone and iPad users for the first time through a standalone Google product rather than a web interface.

The addition of text input addressed one of the most consistent friction points in voice assistant adoption: the social awkwardness of speaking commands in public. Typing support made the Assistant genuinely usable in professional settings. The third major upgrade announced for Assistant was the integration of Google Lens, covered separately below.

Google Home: Hands-Free Calling and Visual Responses

Google Home gained the ability to place voice calls in the United States and Canada at no charge, connecting directly through Google’s infrastructure without requiring a linked smartphone. This put Home in direct competition with Amazon Echo, which had offered calling features earlier that year. The move signaled that smart speaker platforms were shifting from information retrieval devices to communication hubs.

Home also gained the ability to send visual responses to linked screens, including Chromecast, iPhones, and Android devices. The device’s calendar integration was expanded to include proactive updates, such as alerting users to leave earlier for appointments when traffic conditions changed. These updates positioned Home not as a voice search device but as an ambient layer of the user’s daily workflow.

Google Lens: What Is It and How Does It Work?

Google Lens is a visual AI technology that uses machine learning to identify objects, text, locations, and scenes captured by a smartphone camera in real time. CEO Sundar Pichai introduced it at I/O 2017 as a core capability of Google Assistant, not a standalone app. The distinction mattered: Lens was positioned as a persistent intelligence layer, not a feature users had to consciously activate.

The demos shown on stage illustrated three specific use cases: identifying plant species from a camera image, reading and connecting to a home WiFi network by pointing at the router sticker, and pulling restaurant information by pointing at a storefront. A fourth demo, live text translation from Japanese, showed the most immediately practical application for international users. Teams working on AI-native app development quickly recognized Lens as a signal that visual understanding would become a baseline expectation in consumer software.

Google Photos: Shared Libraries and AI-Powered Enhancement

Google Photos had already reached more than 500 million monthly active users by the time of I/O 2017, backing up over 1.2 billion photos and videos each day. The new features announced at the conference focused on social sharing and automatic quality improvement rather than storage expansion.

Suggested Sharing used face recognition to identify people in new photos and automatically prompt users to share images with the people who appeared in them. Shared Libraries allowed family members or groups to maintain a collective photo archive with configurable permissions. Automatic photo enhancement rounded out the update, applying lighting and color corrections without requiring manual editing. Together, these features moved Google Photos from a backup utility into a social product with AI at its core.

Kotlin: Why Google Chose It as Android’s New Language

Kotlin was announced as an officially supported language for Android development at I/O 2017, ending months of community speculation about its status. According to Google’s Android developer documentation, Kotlin is now the preferred language for Android app development. The language had been developed by JetBrains and was already in production use at several major tech companies before Google’s endorsement.

The practical appeal was direct: Kotlin’s null safety, extension functions, and concise syntax reduced boilerplate code dramatically. A block of Java code that required a dozen lines could often be expressed in a single Kotlin statement. For teams evaluating software development trends, Kotlin’s official status meant it was no longer a risk to adopt it was the safer long-term choice for any new Android project.

Virtual Reality: Standalone Headsets and Better Tracking

Source: Google Events

Google Daydream’s standalone VR headset announcement, developed in partnership with Qualcomm, removed the two biggest barriers to VR adoption: the requirement for a tethered smartphone and the requirement for a connected PC. A standalone device meant that the full processing pipeline, display, and tracking happened within a single wearable unit, without external dependencies.

The engineering challenge of standalone VR is significant. Processing visual output, tracking head movement, and rendering a high frame-rate environment within a power-constrained device requires tight hardware-software co-design. The Daydream announcement indicated that Google was investing in this co-design approach rather than simply improving headset aesthetics. For teams building enterprise application development solutions with spatial or simulation components, standalone VR changed the deployment calculus entirely.

TPU: Google’s Hardware Push for AI Workloads

Sundar Pichai announced the second generation of Google’s Tensor Processing Unit (TPU) during the I/O 2017 keynote. The TPU is purpose-built silicon optimized for TensorFlow operations, Google’s open-source deep learning framework. According to Google Cloud’s TPU documentation, these chips are designed specifically to accelerate the matrix multiplication operations that dominate neural network training and inference workloads.

The second-generation chips were made available through Google Cloud, meaning external teams could access TPU compute without building their own hardware infrastructure. For teams working on AI transformation strategy, the Cloud TPU announcement signaled that training large models no longer requires massive on-premise investment. It moved serious machine learning from an infrastructure problem to a software problem, a shift that accelerated the entire field.

Is Your App Ready for AI-First Development?

The shift Google signaled at I/O 2017 is now the baseline expectation for production software. Our engineering team can assess where your current architecture stands and what changes would have the highest impact.

Talk to our team

All Things Android: Studio 3, Instant Apps, and Android Things

Three separate Android developer announcements rounded out I/O 2017 in ways that had immediate practical impact. Android Studio 3.0 added native Kotlin support, a faster Gradle build system, and improved profiling tools for CPU, memory, and network usage. For any team that had been deferring an IDE upgrade, Studio 3.0 gave them clear reason to move.

Android Instant Apps extended its availability to all developers, allowing users to run a specific app function without a full installation. The use case was narrow but commercially valuable: product demos, checkout flows, and feature previews could now reach users who had not committed to installing the full app. Android Things, Google’s IoT-focused operating system, gave development teams a familiar Android API surface for building connected hardware products. Teams interested in how to develop an AI system with hardware components found Android Things a significantly lower barrier to entry than custom embedded development.

What the I/O 2017 Pattern Told Development Teams About Google’s Direction

Looking at the ten announcements together, the strategic logic is consistent: Google was hardening its position as the infrastructure layer for AI-driven software. The Kotlin endorsement reduced friction for Android developers. The TPU gave cloud users access to accelerated ML compute. Google Lens proved that visual AI was production-ready. Android Go extended the addressable market to lower-resource hardware. Each announcement served a different part of the developer ecosystem while reinforcing the same central direction.

For software development in San Diego and across California’s technology sector, I/O 2017 was not an event to watch from a distance. It was a planning input. Teams that restructured their Android builds for Kotlin, started evaluating Cloud TPU for ML workloads, and began designing for Lens integration earlier gained a measurable head start. The companies that waited for full market adoption found themselves backfilling architectural decisions that were avoidable.

The AI workflow automation capabilities that are now standard in enterprise software trace directly back to the infrastructure decisions Google announced at I/O 2017. Understanding that lineage matters for any team evaluating whether their current toolchain is still positioned correctly for the next wave of platform changes.

What We Observed as These Platforms Entered Production Builds

When the Kotlin announcement dropped, our engineering team had already been evaluating it for several projects in progress. The language’s interoperability with existing Java codebases made the transition practical rather than disruptive. What surprised teams most was not the syntax difference but the change in error surface: null safety caught a category of runtime crashes that Java developers had been managing manually through convention rather than enforcement.

The TPU availability through Google Cloud had a more gradual impact. Early access required navigating quota constraints and TensorFlow version compatibility, which meant the practical benefit for smaller teams arrived later than the keynote suggested. Teams in Los Angeles and San Diego who moved quickly on TPU evaluation reported a significant gap between the benchmark performance numbers Google published and the real-world throughput on diverse workloads. The lesson was that AI readiness assessment mattered as much as hardware access. Infrastructure without the right data pipeline and model architecture did not deliver the promised gains.

Building on Google's Platform? Let's Review Your Architecture.

Platform transitions create leverage points for teams that move early. Our engineers can review your current Android, AI, or cloud setup and identify where the highest-value changes are.

Talk to our team

Conclusion

Google I/O 2017 was the conference where the AI-first transition stopped being a strategic statement and became a set of concrete tools, APIs, and platform commitments. Android O expanded what phones could do for users, multitasking across apps. Kotlin reduced the cost of high-quality Android development. Google Lens proved that visual AI was ready for production. The Cloud TPU puts serious machine learning compute within reach of teams without dedicated hardware infrastructure.

The announcements that seemed incremental at the time, Android Go, Instant Apps, and Android Things, turned out to be the ones with the broadest market impact over the following years. Developers who treated I/O 2017 as a roadmap rather than a news event had a clearer view of where to invest their architectural decisions. Understanding Google’s platform direction has always been a core part of building software that stays relevant across platform cycles. The team behind these development decisions is available if you want to discuss where your current stack stands relative to that trajectory.

Frequently Asked Questions