How Google Became the First AI Company

At I/O 2016 and 2017, Sundar Pichai said Google was moving from “mobile first to AI first”. It involves “rethinking all. [its] products and applying machine learning”, with search in images and Smart Answers being early examples. However, for most of the last year, Google was considered to be on the back foot with implementing big language models, and it The AI-first announcement was largely forgotten.

Historically, I/O is where Google makes its biggest announcements of the year. That being said, the AI ​​race has forced the company to make more frequent announcements over the past year, not necessarily at the pace it would have otherwise. With I/O 2024 just days away, here's what we expect Google to announce.

Last year, the stakes of I/O were raised, with many questioning Google's ability to respond to the AI ​​competition. Since then, we've had the Gemini 1.0 and 1.5 family models, as well as various Gemini-branded capabilities in Google's biggest apps. The company is in a better place than it was 12 months ago, but more needs to be done to lead the sector.

To become a truly AI-first company, Google needs to release features that are transformative and widely available, especially to free users.

Apps

The biggest advantage Google has in the AI ​​landscape is that it has millions of users across all of its services that are key to everyday life. While there are many new AI apps coming for early adopters, Google has the ability to introduce more people to AI-powered tools by embedding them in features they already use.

This brings us to Gmail and Docs/Sheets/Slides. Today, Google helps me write, organize and visualize these applications. Those creative AI features are pretty straightforward and solve common tasks that people need to do every day. In terms of what's next, Google previewed the side panel at last I/O and has been testing it in recent months.

I'd argue all of that is a bit on the productivity side, and I'd like to see more personal use cases, especially in Gmail and Calendar, of AI helping people in other aspects of their lives.

Other key apps that people use are Google Maps and Search. The Search Generative Experiment (SGE) was announced a year ago and I wonder if Google considered it mature enough to move out of the Labs preview program. I can see the utility of providing a direct answer instead of requiring people to search through links. At the same time, the implications of this approach for publishers are – to say the least – profound.

Google Maps is testing creative AI search to let you find places in a conversational way. Compared to web search, I find conversational search is better in a more limited domain.

Then there is Gemini, the application and the website. On mobile, the functionality is lacking and it's not really a good phone assistant, at least not to the extent of the long-established Google Assistant. It certainly looks like the upcoming updates will finally figure it out.

Chrome is another major application that people use. It just added a Gemini shortcut to the address bar which is very handy. Historically, we don't really see the browser discussed at I/O. I'm curious about Google's full vision of what AI looks like in Chrome. So far, it's introduced Help Me Write, a tab organizer, and a theme generator in recent months, but nothing significant has changed yet, especially compared to competition that's skewed towards AI.

Platforms

Then there is the platform on which people are using these apps. Google has detailed what's coming in Android 15. How big these updates are after last year's surprisingly limited stage presence and quiet public release is a different question.

The primary use of generative AI in Android today is the Gemini Nano powering Gboard Smart Replies, Messages Magic Compose, and Summary Recorder. On-device AI will be critical to helping keep cloud costs down, and making Android the place to build gen AI mobile applications.

That said, save for wallpapers, we haven't really seen AI applied to the user experience. Rethinking the OS for AI is a big question. For example, what does a smart home screen, lock screen, or notification shade look like? What can you do with macros, shortcuts in iOS or third-party Tasker on Android, and how can Google replicate and automate it with AI?

Comparisons will be made when Apple launches iOS 18 later this year, which is expected to include a host of on-device AI features. Of course, the big difference is that Android and its apps can be updated more easily than Apple's monolithic OS.

On the platform front, there's Wear OS 5 and (probably) the long-running Android XR. While Meta is working to make Horizon OS available to other headset makers, Google's Samsung partnership is still toiling away in the background.

Hardware

Moving up the second layer is the hardware that is running these OS and apps. Google unexpectedly announced the Pixel 8a as well as a standalone Pixel tablet a week before I/O. Officially, this was to give 8a its own moment. The mid-ranger would have been lost in a flurry of other I/O news.

The question now is whether Google will tease the Pixel 9 series at I/O 2024. This happened for the Pixel 7 but not the Pixel 8. One argument for not being the case is that the Pixel 8 and 8 Pro are still solid phones with some. For the runaway example, we're still waiting for Zoom Enhance to launch and it would be odd if Google announced the next phone in any capacity before the Pixel 8's feature set is complete. More anecdotally, in the US, Google is still heavily promoting the 8 Series as part of the NBA playoffs.

What I can see Google teasing is the rumored “Pixie” AI assistant. A precedent here would be how Google showed off the new Google Assistant at I/O several months before the Pixel 4 launch. This may help reduce overstimulation, although it may not be available on current devices.