According to The New York Times, Apple has hired John Giannandrea, Google’s chief of search and artificial intelligence. In a memo obtained by The Times, Tim Cook said:

“Our technology must be infused with the values we all hold dear,” Mr. Cook said in an email to staff members obtained by The New York Times. “John shares our commitment to privacy and our thoughtful approach as we make computers even smarter and more personal.”

Giannandrea joined Google in 2010 as part of the company’s acquisition of Metaweb and is credited with infusing artificial intelligence across Google’s product line. Giannandrea will report directly to Apple CEO Tim Cook.

This is a huge ‘get’ for Apple and comes fast on the heels of reports that the company is hiring over 100 engineers to improve Siri.

Nick Heer at Pixel Envy tested how well 2018 Siri performs commands given to the voice assistant in a 2010 demo video. The video takes Siri, which started as a stand-alone, third-party app, through a series of requests like ‘I’d like a romantic place for Italian food near my office.’ Just a couple of months after the video was published, Siri was acquired by Apple and the team behind it, including the video’s narrator, Tom Gruber, began integrating Siri into iOS.

That was eight years ago. Inspired by a tweet, Heer tested how well Siri performs when given the same commands today. As Heer acknowledges, the results will vary depending on your location, and the test is by no means comprehensive, but Siri’s performance is an eye-opener nonetheless.

What’s clear to me is that the Siri of eight years ago was, in some circumstances, more capable than the Siri of today. That could simply be because the demo video was created in Silicon Valley, and things tend to perform better there than almost anywhere else. But it’s been eight years since that was created, and over seven since Siri was integrated into the iPhone. One would think that it should be at least as capable as it was when Apple bought it.

Eight years is an eternity in the tech world. Siri has been fairly criticized recently for gaps in the domains it supports and their balkanization across different platforms, but Heer’s tests are a reminder that Siri still has plenty of room for improvement in how it handles existing domains too. Of course, Siri can do things in 2018 that it couldn’t in 2010, but it still struggles with requests that require an understanding of contexts like location or the user’s last command.

Voice controlled assistants have become a highly competitive space. Apple was one of the first to recognize their potential with its purchase of Siri, but the company has allowed competitors like Amazon and Google catch up and pass it in many respects. The issues with Siri aren’t new, but that’s the heart of the problem. Given the current competitive landscape, 2018 feels like a crucial year for Apple to sort out Siri’s long-standing limitations.

Bryan Irace writes about one of the biggest challenges Apple faces with Siri:

It’s no easy task for a voice assistant to win over new users in 2018, despite having improved quite a great deal in recent years. These assistants can be delightful and freeing when they work well, but when they don’t, they have a tendency to make users feel embarrassed and frustrated in a way that GUI software rarely does. If one of your first voice experiences doesn’t go the way you expected it to – especially in front of other people – who could blame you for reverting back to more comfortable methods of interaction? Already facing this fundamental challenge, Apple is not doing themselves any favors by layering on the additional cognitive overhead of a heavily fragmented Siri experience.

I think Irace is right on in this observation – Siri’s fragmentation is a real problem.

On the more optimistic side, it could be taken as good news that the fix appears fairly obvious: create a single Siri that’s consistent across all platforms. This seems like it would be a clear net positive, even though such a change could reduce Siri’s accuracy in some cases; for example, I’m guessing Siri on the Apple TV is currently tuned to expect TV and movie queries more than anything else, so it can more effectively produce the right kind of results – tweak that tuning, and Apple will have to work even harder at helping Siri understand context.

One thing that’s concerning about the apparent simplicity of this fix is that Apple hasn’t made it yet, meaning, perhaps, that the company thinks there’s nothing wrong with Siri’s current fragmentation. This conversation would be different entirely if Apple had begun showing an increased effort to unify Siri across its platforms, but recently, the opposite has been true instead. The latest major Apple product, HomePod, includes a stripped-down Siri that can’t even handle calendar requests. And SiriKit, which launched less than two years ago, was designed in a way that fundamentally increases fragmentation. Irace remarks:

If the Lyft app is installed on your iPhone, you can ask Phone Siri to order you a car. But you can’t ask Mac Siri to do the same, because she doesn’t know what Lyft is. Compare and contrast this with the SDKs for Alexa and the Google Assistant – they each run third-party software server-side, such that installing the Lyft Alexa “skill” once gives Alexa the ability to summon a ride regardless of if you’re talking to her on an Echo in your bedroom, a different Echo in your living room, or via the Alexa app on your phone.

The only recent occasion that comes to mind when Siri has moved in the right direction – gaining knowledge on one platform that previously existed only on another – was when iOS 10.2 brought the full wealth of Apple TV Siri’s movie and TV expertise to iOS. This only happened, though, because iOS 10.2 introduced the TV app.

Until Siri can answer the same requests regardless of what platform you’re on, most people simply won’t learn to trust it. Users shouldn’t have to remember which device’s Siri can answer which questions – all they should have to remember is those two key words: “Hey Siri.”