Google started as a search engine, as we put 2017 behind us, it's all the more clear just how far the Mountain View company has come. Android has grown into the most popular computing platform in the world, and Google Assistant is running on a huge number of those devices. Machine learning backs so much of what Google did in 2017, making it clear this is no longer about impressive hypothetical research. Machine learning is the future of Google, and it's having huge impacts on the way we use technology right now.
Oreos on the "Go"
No discussion of Google's efforts in 2017 can ignore the continued importance of Android. In 2017, Android surpassed Windows as the most popular operating system on Earth. People who buy smartphones are much more likely to buy one running Android than iOS when you look at the global numbers, and people buy a lot of smartphones.
Google is looking toward the future with Android, as well, In 2017, Google announced the Android Go platform, a stripped down version of Android designed to run on ultra-budget phones with limited storage and processing power. Android Go, which will be a variant of Android 8.1, even has its own suite of apps like Maps and Gmail that run smoother and use less data.
In many places, smartphone usage has reached saturation. Many of us still pick up new devices every year or two, but there are many places where smartphone usage is still picking up steam. Google wants Android to be running on the "next billion" smartphones, and Android Go is how we get there.
Of course, it's not all about the entry-level phones. Google also released Android 8.0 and 8.1 Oreo in the fall after starting a developer preview in spring 2017. It was a little disappointing Google didn't do a big promotional push for Oreo like it did for KitKat a few years back, but aside from the awkward launch, Oreo is a good update.
There aren't as many headlining features in Oreo as in some past version of Android, but the under-the-hood improvements will make a big impact going forward. Google is cracking down on background processes to improve battery life, and users now have more control over how apps push notifications. There's also Project Treble, which aims to solve the problem of fragmentation once and for all. This modular system framework will run on all phones that ship with Oreo, allowing OEMs to make system updates that don't require new hardware drivers. That means faster updates and longer support.
Google launched Assistant at the very tail-end of 2016, but the platform truly came into its own in 2017. It runs on all recent Android phones now. The rollout started in early 2017 with phones running Android 6.0 and higher, and recently it expanded to include Lollipop phones and Nougat tablets. Billions of people have Assistant in their phones, and it's vastly more useful than Google Now on Tap ever was.
Assistant doesn't only live in phones—it's part of Google Home as well, and Google has sold mountains of these speakers. There's the original Home, as well as the super-cheap Home Mini and the new high-end Home Max that launched in late 2017. Google has leveraged the power of its search database and machine learning to make Assistant smarter than all the other virtual assistants on the market. However, Google continues to lag behind Amazon in product integrations. Assistant is rapidly making up ground, though.
Google the hardware company
Google used to only release hardware that had someone else's name on it too. For example, all those Nexus phones and tablets of the years that it made with hardware partners like LG and Huawei. Now, Google has its own hardware team under the direction of former Motorola CEO Rick Osterloh. This has become an important part of Google's business, but not all its products are good.
The Pixel Buds were a flop despite an impressive first showing with its real-time translation demo. However, the expensive earbuds are buggy, don't fit well, and lack the futuristic completely wireless design of other products. Meanwhile, the Pixelbook has gotten rave reviews for the most part, but the price tag makes it a tough sell.
The aforementioned Google Home speakers are important to Google's Assistant rollout, but they're also very nice pieces of hardware. The Home Max is really expensive, but it's a seriously impressive speaker. Meanwhile, the Home Mini is so cheap it's basically an impulse purchase. Like a lot of Google's hardware, these devices have lots of curves and fabric patches to make them feel more approachable and homey.
2017 also brought us the second generation of Pixel phones. Google fixed a number of issues from last year, but it also removed the headphone jack. The Pixel 2 XL OLED display was also lacking compared to many other phones. Google's decision to have LG build that phone with its own OLED tech resulted in a lot of bad press for Google, but at the end of the day, these are still the best Android phones you can get. If the XL's screen bugs you, there's always the smaller Pixel 2 with all the same features.
Google might be relying on its hardware partners even less going forward. There were rumors in 2017 that Google would acquire HTC, but in the end it only bought HTC's Pixel division. That's a few thousand designers and engineers joining the hardware team to make a new generation of Google hardware.
Google's machine learning future
The annual Google I/O is a chance to find out where Google is directing its energies, and there was one clear overarching theme this year: machine learning. Google has been talking about machine learning and artificial neural networks for years, but it's only recently that we've been benefiting directly as users of Google services. If you've ever entered a search term in Google Photos, you're taking advantage of Google's incredible expertise in machine learning.
At I/O, Google announced a new generation of its Tensor Processing Units, a custom chip that runs neural network operations. Each second-gen TPU (see below) can deliver 180 teraflops of performance. Some of these chips are scanning your photos to identify sunsets, dogs, and even which people are visible. In one particularly interesting feat of machine learning prowess, Google Photos gained the ability to automatically share images of certain people with one of your contacts. For example, making sure you and your spouse both have all the snapshots of the kids.
One of the other machine learning innovations shown off at I/O has only started appearing on devices. Google Lens was demoed as a way to get contextual data about the world around you based on a photo. It launched on the Pixel 2 devices in the Photos app, and later rolled out to All Pixels in Assistant. You simply show Lens an object, and it figures out what it is. It's still hit and miss, but it can do things like find reviews for a restaurant when you take a picture of the sign or save a business card's contact info to your contacts. When these things work well, it's really impressive.
You might have noticed Google's Assistant voice got a makeover in 2017 as well. That's thanks to a text-to-speech platform called Wavenet, which comes from Google's DeepMind machine learning subsidiary. No one else is anywhere close to Google when it comes to speech synthesis, and it's still pushing the envelope. An upcoming revision of its text to speech technology is reportedly indistinguishable from a person. I remain skeptical until I hear it in practice, but it seems feasible.
Moving forward in 2018
Google will have some big decisions to make in 2018 that can't be solved with machine learning (well, probably). The company hasn't released an Android tablet for two years now, and no one else is making compelling and successful Android slates, either. At the same time, Chrome OS has become a mature and capable computing platform. The solution, perhaps, is a Chrome OS tablet.
The Pixelbook is almost a replacement for Android tablets with its form factor and Android app support. There are still places Chrome OS doesn't work well with touch, and that's something Google needs to address if it's going to sell a tablet ever again.
Android Wear has also stagnated with several prominent smartphone makers calling it quits. It's basically all fashion retailers pushing new Wear devices, none of which are very compelling. Google hasn't helped matters with its clunky decisions in Wear 2.0. Google needs to do some serious work in 2018 on Wear if it wants to keep selling smartwatches.
We're also on the verge of a potentially major shift in the way we interact with ads. In the coming months, Google plans to roll out ad-blocking in Chrome. It won't block all ads, but websites that are flagged as having intrusive or obnoxious ads will be stripped bare. Chrome is the most popular browser in the world, so this could finally push everyone to run less terrible ads. Look for that to happen later in the spring.
So, that's Google's past, present, and future. It's been a wild 2017, and 2018 will no doubt be even more exciting.