Tech
Chatbot looking out of a smartphone display. Text bubbles floating around. Pink background.
Getty Images

Amazon and Apple are struggling with their AI voice assistants

Two separate pieces on the AI assistants suggest good versions are a ways off.

Rani Molla

As we wrote last year, voice assistants have seemingly gotten worse as Big Tech companies try to transition the technology underlying them to compete in a ChatGPT world. The idea is to supercharge Apple’s Siri and Amazon’s Alexa with AI so they can finally become true assistants, advancing beyond just reliably playing music, setting timers, and telling you the weather.

The problem lies in the switch from natural language processing — a rigid but consistent system that detects what you’re trying to say using preprogrammed intent models — to large language models, which are more expansive and generate probable answers by analyzing vast amounts of text. In the process, something seems to have broken, and these tools are now often worse at the basic tasks they used to handle well.

It turns out, things aren’t really getting better with time. Behold: two recent pieces on the state of Apple and Amazon’s voice assistants:

The New York Times’ Kevin Roose:

“The good news is that the new Alexa+ is, in fact, more fun to talk to than the old one, with more realistic synthetic voices and a more humanlike cadence. (There are eight voices to choose from; I used the default setting, an upbeat female voice.)

And I liked some of Alexa+’s new capabilities, such as booking a table at a restaurant and generating long stories and reading them to my 3-year-old.

The new Alexa is also better at handling multistep requests. ‘Set three kitchen timers for 15, 25 and 45 minutes’ and ‘write a one-day itinerary for a trip to San Diego and send it to my email’ were two prompts that worked for me. And Alexa+ doesn’t require you to say its wake word every time you talk to it, so you can go back and forth or ask it follow-up questions, which is a nice change. The bad news is that despite its new capabilities, Alexa+ is too buggy and unreliable for me to recommend. In my testing, it not only lagged behind ChatGPT’s voice mode and other A.I. voice assistants I’ve tried, but was noticeably worse than the original Alexa at some basic tasks. When I asked Alexa+ to cancel an alarm the other morning — a request I had made to the old Alexa hundreds of times with no issues — it simply ignored me.

When I emailed a research paper to alexa@alexa.com, in order to hear Alexa+ summarize it while I washed the dishes, I got an error message saying the document couldn’t be found.

Alexa+ also hallucinated some facts and made some inexplicable errors. When I asked it to look up Wirecutter’s recommended box grater and add it to my Amazon cart, it responded that ‘according to Wirecutter, the best box grater is the OXO Good Grips Box Grater.’ Wirecutter’s actual box grater pick is the Cuisipro 4-Sided Box Grater. Luckily, I caught the mistake before ordering. When I asked Alexa+ to walk me through installing a new A.I. model on my laptop, it got tripped up and started repeating, ‘Oh, no, my wires got crossed.’”

Bloomberg’s Mark Gurman:

“The plan now is to ship [Apple Intents, a tool that lets you voice operate your iPhone with precision] alongside a broader Siri infrastructure overhaul in the spring and market it heavily. But there’s some concern inside the company, I’m told. Engineers have been struggling to ensure that the system works with a sufficient number of apps and is accurate enough to handle high-stakes scenarios. There are worries about the software failing in categories where precision is nonnegotiable, like in health or banking apps.

For years, users have struggled with Siri not understanding them. It’s annoying but not critical if your phone misunderstands the city you want a weather report from or tries to navigate you to the wrong restaurant. But letting the Siri brain of today control all of your apps would obviously be a lot riskier.

That’s why Apple is waiting on the new Siri and won’t roll it out universally on day one. Testing is underway with select third-party apps, including Uber, AllTrails, Threads, Temu, Amazon, YouTube, Facebook, WhatsApp, and even a few games, in addition to Apple’s own apps. For banking and other sensitive categories, Apple is considering sharply limiting what Siri can do — or excluding those areas altogether.

This isn’t just about making Siri smarter. It’s about giving Apple’s ecosystem a new, voice-first interface. If the company is actually able to bring it to market (and that’s a gigantic if), it could potentially be a hit that many users didn’t see coming.”

So Amazon’s AI voice assistant hasn’t been able to balance its new talents while performing its old ones, and it’s buggy all around. And Apple doesn’t expect its AI assistant to come out until spring 2025, as it works out problems internally.

More Tech

See all Tech
tech

Apple stock takes a hit on report it’s pushing back AI Siri features — again

Apple customers may have to wait even longer for the company’s long-awaited AI Siri, Bloomberg reports.

The iPhone maker had been planning to include a number of upgrades to Siri in a March operating system update, but the company now is planning to spread those out over future versions. That means some features first announced in June 2024 — an AI Siri that can tap into personal data and on-screen content — might not arrive until September with iOS 27.

The postponements happened after “testing uncovered fresh problems with the software,” Bloomberg said, including instances where Siri didn’t properly process queries or took too long to respond.

The stock, which had been trading up more than 2% today, has pared some of those gains on the news.

For what it’s worth, Apple’s iPhone sales — a record last quarter — don’t appear to be suffering for lack of AI.

The postponements happened after “testing uncovered fresh problems with the software,” Bloomberg said, including instances where Siri didn’t properly process queries or took too long to respond.

The stock, which had been trading up more than 2% today, has pared some of those gains on the news.

For what it’s worth, Apple’s iPhone sales — a record last quarter — don’t appear to be suffering for lack of AI.

tech

Meta breaks ground on massive $10 billion AI data center — and the costs won’t stop there

Meta announced today that it broke ground on a new, giant AI data center. This one is located in Indiana, has 1 gigawatt of capacity, and will cost more than $10 billion.

In a press release, the company touted the 4,000 construction jobs and 300 operational positions Meta expects to bring to the area. It did not disclose any tax incentives tied to the project.

But much like with the company’s Hyperion data center in Louisiana — where we calculated incentives in the billions — the number of long-term jobs is likely small relative to any public subsidies the company ultimately receives.

The $10 billion build represents a notable chunk of Meta’s planned $115 billion to $135 billion in capital expenditure this year. And operating costs will add substantially to that total over time.

Earlier this year, President Trump warned tech giants to “pay their own way” when it comes to energy, as data centers have driven up electricity costs in some regions. Meta’s announcement appears to anticipate that criticism, dedicating significant space to explaining how it will mitigate the energy and water impact of the facility:

“With all our data centers, we strive to be good neighbors. We pay the full costs for energy used by our data centers and work closely with utilities to plan for our energy needs years in advance to ensure residents aren’t negatively impacted. To help support local families in need, we’re providing $1 million each year for 20 years to the Boone REMC Community Fund to provide direct assistance with energy bills, and funding emergency water utility assistance through The Caring Center. We also pay the full cost of water and wastewater service required to support our data centers. Over the course of this project, Meta will make investments of more than $120 million, toward critical water infrastructure in Lebanon, as well as other public infrastructure improvements including roads, transmission lines and utility upgrades.”

Unlike hyperscalers such as Google and Microsoft, which can offset infrastructure costs by selling cloud capacity to customers, Meta bears those expenses largely on its own. That dynamic could make the economics of AI infrastructure more challenging for the company as its AI spending continues to accelerate.

But much like with the company’s Hyperion data center in Louisiana — where we calculated incentives in the billions — the number of long-term jobs is likely small relative to any public subsidies the company ultimately receives.

The $10 billion build represents a notable chunk of Meta’s planned $115 billion to $135 billion in capital expenditure this year. And operating costs will add substantially to that total over time.

Earlier this year, President Trump warned tech giants to “pay their own way” when it comes to energy, as data centers have driven up electricity costs in some regions. Meta’s announcement appears to anticipate that criticism, dedicating significant space to explaining how it will mitigate the energy and water impact of the facility:

“With all our data centers, we strive to be good neighbors. We pay the full costs for energy used by our data centers and work closely with utilities to plan for our energy needs years in advance to ensure residents aren’t negatively impacted. To help support local families in need, we’re providing $1 million each year for 20 years to the Boone REMC Community Fund to provide direct assistance with energy bills, and funding emergency water utility assistance through The Caring Center. We also pay the full cost of water and wastewater service required to support our data centers. Over the course of this project, Meta will make investments of more than $120 million, toward critical water infrastructure in Lebanon, as well as other public infrastructure improvements including roads, transmission lines and utility upgrades.”

Unlike hyperscalers such as Google and Microsoft, which can offset infrastructure costs by selling cloud capacity to customers, Meta bears those expenses largely on its own. That dynamic could make the economics of AI infrastructure more challenging for the company as its AI spending continues to accelerate.

tech

Humanoid robot maker Apptronik raises $520 million

Apptronik, an Austin, Texas-based robot manufacturer, said it has closed out its Series A fundraising round, raising $520 million. The fundraising is an extension of a $415 million round raised last February, and included investments from Google, Mercedes-Benz, AT&T, and John Deere. Qatar’s state investment firm, QIA, also participated in the fundraising round.

Apptronik makes Apollo, a humanoid robot targeted for warehouse and manufacturing work. The company is one of several US robotics companies that are racing to apply generative-AI breakthroughs to humanoid robots, in anticipation of a new market for robots in homes and workplaces.

Apptronik makes Apollo, a humanoid robot targeted for warehouse and manufacturing work. The company is one of several US robotics companies that are racing to apply generative-AI breakthroughs to humanoid robots, in anticipation of a new market for robots in homes and workplaces.

tech

Ives: Microsoft and Google’s giant capex plans are worth it

Don’t mind the AI sell-off, says Wedbush Securities analyst Dan Ives, who thinks fears around seemingly unfettered Big Tech capex budgets are unfounded, especially in the case of Microsoft and Google. Together, the two hyperscalers are slated to spend around $300 billion on the purchases of property and equipment this year as they double down on AI infrastructure, but he says both have already shown that they can turn the spending into revenue and growth.

“They are reshaping cloud economics around AI-first workloads that carry higher switching costs, deeper customer lock-in, and longer contract durations than before,” Ives wrote, adding that these giant costs will be spread out over time and set the companies up for success in the long run. Per Ives:

“While near-term free cash flow optics remain noisy, the platforms that invest early and at scale are best positioned to capture durable share, pricing power, and ecosystem control as AI workloads mature. Over time, we expect utilization leverage to turn today’s elevated investment into a meaningful driver of long-term value creation.”

“They are reshaping cloud economics around AI-first workloads that carry higher switching costs, deeper customer lock-in, and longer contract durations than before,” Ives wrote, adding that these giant costs will be spread out over time and set the companies up for success in the long run. Per Ives:

“While near-term free cash flow optics remain noisy, the platforms that invest early and at scale are best positioned to capture durable share, pricing power, and ecosystem control as AI workloads mature. Over time, we expect utilization leverage to turn today’s elevated investment into a meaningful driver of long-term value creation.”

Latest Stories

Sherwood Media, LLC produces fresh and unique perspectives on topical financial news and is a fully owned subsidiary of Robinhood Markets, Inc., and any views expressed here do not necessarily reflect the views of any other Robinhood affiliate, including Robinhood Markets, Inc., Robinhood Financial LLC, Robinhood Securities, LLC, Robinhood Crypto, LLC, or Robinhood Money, LLC.