Innovation, AI and Google I/O: A 2016 Discussion

In a conversation originally published in May 2016, Kyle Russell, Benedict Evans, Connie Chan and Steven Sinofsky react to Google I/O. They discuss AI assistants, messaging, Android Instant Apps and virtual reality as announcements and predictions of their time. [Source: a16z](https://a16z.com/podcasts/a16z-show/).

Russell notes that critics saw familiar ideas in voice assistants and messaging bots. Sinofsky responds that innovation often consists of combining and refining existing ideas, rather than an invention appearing fully formed.

A new product can look much like its predecessors while changing the experience through a different combination of capabilities. Sinofsky uses smartphones and messaging apps to argue against judging innovation solely by novelty.

The participants see artificial intelligence as a theme connecting Google’s announcements. Evans describes a progression from indexing pages to understanding entities and eventually anticipating what a user might need.

Chan says the messaging demonstration illustrates reducing friction. The group imagines how an assistant might use calendars, email and context to make everyday tasks easier, while acknowledging that automated suggestions can be wrong.

The speakers compare early AI mistakes with earlier navigation, spelling and autocorrect errors. They expect the tools to improve over time; this is a forecast made at the event, not proof that every risk will disappear.

Russell lists announcements including the Allo messaging app, Google Assistant and Instant Apps. Evans distinguishes individual product tactics from the broader strategy of using data and AI across many services.

Evans sees the assistant as an attempt to move from a list of search links toward answers and suggestions in context. Sinofsky emphasizes the distinction between an individual app and the platform beneath it.

On messaging, the group argues that the market was still fragmented and that existing apps left room for improvement. Their predictions about future consolidation and Google’s products were contemporary speculation.

The discussion turns to Instant Apps. The announced approach let an Android user follow a link and load relevant parts of an app without first going through the conventional full-install flow.

Evans describes how app links previously opened installed apps or fell back to a web page. Instant Apps sought to soften that boundary, raising questions about discovery, device storage and how developers structure apps.

Sinofsky frames the change as another attempt to deliver code to a device using the computing power and bandwidth available at the time. The web and native apps make different tradeoffs in reach, capabilities and deployment.

Chan compares lower-friction interactions with lessons from WeChat, particularly reducing repeated sign-ins and payment entry. The speakers contrast native app delivery with browser views and conversational bots as different ways to reach users.

Russell raises Google’s VR announcements: Cardboard had put simple phone-based VR into many hands, while Daydream proposed a more capable Android platform and reference design for partners.

Evans predicts that affordable mobile hardware could make VR more accessible than a high-cost computer-tethered setup. This was his 2016 prediction about a developing market, not a statement of what later happened.

The group compares Google’s partner ecosystem with more vertically integrated approaches. Chan warns that a range of hardware partners can also mean inconsistent quality and a more complicated experience for users and developers.

Sinofsky expects experimentation and uneven quality as companies decide which layers to standardize and which to differentiate. The discussion closes with the participants’ enthusiasm for watching those approaches develop.