The AI Will Come Out Tomorrow
There was a time, not too long ago, that Alexa and Siri were battling each other at the vanguard of so-called voice assistants. It's a space where Apple was an actual first-mover, for once (thanks to an acquisition done under Steve Jobs). But Amazon quickly zoomed ahead in the race thanks to a strategy that was basically the opposite of Apple's: cheap Echo devices, everywhere. Still, thanks to being baked into every iPhone, both Alexa and Siri were in the hands and homes of hundreds of millions of users. These days they find themselves battling in a new way: who can actually ship upgraded versions of their assistants in the age of AI without embarrassing themselves...
I would say "the race is on" except that the race actually keeps getting delayed.
In posts that were oddly, but seemingly coincidentally published within minutes of one another on Friday, Caroline O'Donovan wrote about the new version of Alexa being delayed, again, for The Washington Post. Meanwhile, Mark Gurman got the scoop that the new Siri also faces "possible" delays as well, for Bloomberg.
Those who have been following the "Remarkable" Alexa saga probably could have seen this coming. It was previously reported that Valentine's Day would be the go/no-go date to make the call if the new version was ready to ship. Alexa, it seems, saw her shadow. She will not be coming early:
The “smarter and more conversational” version of Alexa will not be available until March 31 or later, the employee said, at least a year and a half after it was initially announced in response to competition from OpenAI’s ChatGPT.
Internal messages seen by The Post confirmed the launch was originally scheduled for this month but was subsequently moved to the end of March.
Sure, the end of March doesn't sound too bad as far as delays go, but the problem is that the event is already in the books for February 26. Rather than saying that the new Alexa will be available at the event, they'll have to give some sort of "coming soon" message. And the problem there is that, as noted, they've been saying that for the past 18 months. But delay after delay after delay has kept this new Alexa quiet.
It's embarrassing for Amazon – but presumably not as embarrassing as launching a version of Alexa that doesn't work or underwhelms. I'm still pretty worried about both scenarios just given these delays and all the various reports over the past many months. But even without those, what Amazon – and Apple – are trying to do is extremely hard. As I wrote on this topic last year, "Being Too Early Is Worse Than Being Late" – in ways, this is actually more difficult than trying to launch a new AI assistant from scratch. Because they have so many legacy users of their previous products, there are expectations and undoubtedly gnarly backwards-compatibility issues. Amazon doesn't want to become Sonos.
Nor does Apple! Siri probably has more leeway to make major changes because she's been viewed largely as a joke. Harsh, but true! After launching as sort of a novelty with a lot promised, instead over the years we've been promised that this was the update which would make Siri actually good. Has she gotten better? I mean sure, in ways. But the reality remains that she gets far, far, far, far too much wrong to be considered reliable enough to trust beyond perhaps setting timers and playing music.
And that matters more for Siri because Apple is touting her as the one such assistant you can trust with your personal data, thanks to Apple's security practices. That may or may not be true, but I also know there's no way I'm trusting Siri to do something important until she proves herself. Over and over again. Over a long period of time. And so news of delays doesn't give a lot of confidence:
Apple hinted that the changes wouldn’t be released until 2025, but it wasn’t yet clear how long they would take to arrive. Internally, the company has been planning to introduce the technology as part of iOS 18.4, the version of its mobile operating system coming in April.
But now Apple is considering delaying or limiting at least some of the overhaul until iOS 18.5, which will be released as early as May, the people said.
Gurman does couch this information a bit. First and foremost, as he notes, Apple has never actually said when these new "trusted" Siri features would ship (despite showing them off last June at WWDC), and they could still technically ship them in the iOS 18.4 beta, just not enabled. But sorry, Apple gets no benefit of the doubt with regard to Siri. Does anyone doubt that they're having a hard time getting this new version of Siri to live up to her promise? No.
And part of it is the same issues that Amazon is facing: because of their previous "success" (in rolling out Siri far and wide, at least), they're more now handcuffed than if they were started from scratch. Google ran into this issue as well with the launch of Gemini on Pixel devices last year, and they made users choose whether to keep using the old 'Assistant' versus "upgrading" to Gemini, noting that some old features, such as yes, setting timers, wouldn't be available at first.
But again, Google didn't have the same ubiquity problem as Apple or Amazon. Sure, there are billions of Android devices, but the assistant features are fragmented depending on the phone maker. And the Gemini roll-out was limited to start. Apple and Amazon seem to be planning to go far and wide.1
And, of course, this isn't just about legacy users and features, Amazon and Apple are now (re)entering a market where OpenAI has changed the game with ChatGPT. LLM-based assistants make the older assistants seem stupid, quite literally. And now the market has Anthropic's Claude, Mistral's Le Chat, Perplexity, the aforementioned Gemini – even DeepSeek!
Again, the risk of embarrassment is high here for Apple and Amazon. As bad as it is to have to keep pushing back these roll-outs, it's not nearly as bad as rolling them out only for them not to work.
To that end, I continue to wonder if it wouldn't make more sense for Apple to temporarily outsource Siri to ChatGPT. I know this sounds extreme, but I'm really not sure it's the worst idea in the world, especially given the partnership with OpenAI already in place. You could keep Siri powering the simple stuff like timers and music, but just default all "world knowledge" queries to ChatGPT. Obviously, users would have to opt-in, but if I were Apple, and Siri really isn't ready to roll, I would definitely consider this option.
Same with Amazon given their massive investment in Anthropic. What if Alexa keeps powering timers and Claude handles most everything else? Various reports indicate this may be at least part of how the new Alexa works anyway. The problem with fully outsourcing until Alexa is ready for prime time may be the notion that Amazon is trying to charge users for this new Alexa. That's another added layer of complexity (and perhaps embarrassment) here. It's one thing if the free service baked into your device doesn't work. It's another if you buy a new device to use a service you have to also pay monthly for. There's a lot of risk for Amazon here.
All of this leads me to believe that both of these companies are probably going to be in the market this year to acquire teams/products to help their AI pushes.
Some reports now indicate that Amazon is starting to right the ship with their own internal AI work (though, it's perhaps extra complicated by many cooks in the kitchen) and that's good news since Anthropic may now be outside of the acquisition window (presuming Amazon even could do such a deal in the new regulatory environment) with their new round valuing them at $60B. Apple may be more primed to scoop someone up, but historically shies away from big deals. Still, to save Siri – again, one of their higher profile acquisitions from the Jobs-era – it may be warranted.
First, we'll see how Alexa and Siri perform post-launch. Assuming they launch. Tomorrow is always a day away, it seems.
M.G. Siegler
M.G. Siegler
M.G. Siegler
1 There had been some talk that Amazon may keep a "legacy" version of Alexa around -- in particular for free users on older devices -- but it's not clear if that will actually be the strategy.