• | 9:00 am

AI’s giants are still demoing stuff faster than they can ship it

As OpenAI, Google, and Microsoft flood the zone with AI announcements, telling what’s real is only getting tougher.

AI’s giants are still demoing stuff faster than they can ship it
[Source photo: Microsoft CEO Satya Nadella speaks during an event at Microsoft headquarters in Redmond, Washington, on May 20, 2024. [Photo: JASON REDMOND/AFP via Getty Images]]

A new AI era begins

Signs displaying that confident declaration were plastered all over Microsoft’s campus when I visited on Monday. I wasn’t sure if it referred just to the Windows-centric AI media briefing the company was holding or also to its Build developer conference, which was starting the following day. But coming the week after OpenAI’s unveiling of its ChatGPT Voice Mode and Google’s avalanche of AI announcements at its I/O conference, my instinctive reaction was: Another new AI era? So soon?

With all due respect to Microsoft, its news—which included new Copilot+ PCs, based on Qualcomm’s first serious PC processors, and incorporating AI-infused features such as Recall search—didn’t usher in an AI era all by itself. Still, with three of AI’s giants taking center stage in such quick succession, the last week and a half has been remarkably newsy. Another big shoe will drop on June 10, when Apple will likely spell out a more AI-heavy vision for the iPhone, Mac, iPad, and other devices at its WWDC keynote.

Among Microsoft’s announcements at Build: New features for its Copilot Studio platform, coming sometime later this calendar year as a public preview. [Image: courtesy of Microsoft]

That’s not even counting the recent AI news that refuses to cooperate with carefully orchestrated product rollouts—most notably the mess OpenAI got itself into by giving ChatGPT Voice Mode a default voice that’s like a tribute-band version of Scarlett Johansson’s performance in Her, much to Johansson’s displeasure. (OpenAI says it’s “working to pause” use of the voice; I suspect the kerfuffle—which many people are taking as a bigger story about the AI industry’s ransacking of human talent for its own purposes—is far from over.)

So much stuff is happening so quickly that trying to devote equal mindshare to all of it would make your head explode. Google said it mentioned AI 121 times during its I/O keynote, and Microsoft’s official list of its announcements at Build included 59 discrete items. But one big reason why the current news cycle is so intense is that it has very little to do with what’s ready to ship right now. Instead, as the companies in question jostle for AI leadership, they’re showing off vast quantities of works in progress, sometimes without being clear about how close to reality they are.

Was it sheer coincidence that OpenAI revealed its new GPT-4o model and Voice Mode the day before Google I/O kicked off? It doesn’t matter. Either way, its demos—including the not-Scarlett-Johansson voice offering grooming advice for a job interview and fussing over a puppy—were dazzling. And they sucked up a meaningful percentage of the attention that would have otherwise gone to the Google event.

We do know for sure that OpenAI announced Voice Mode before it was ready to start deploying it to users. We’ll start seeing it “soon,” the company says, and until we do, it’s impossible to gauge just how useful it will be in the real world. That’s a strikingly different situation than with OpenAI’s November 2022 reveal of the original version of ChatGPT: It was available the moment the company acknowledged its existence with the simple invitation, “Try it now.”

Soon might be the single most commonly stated shipping date for AI products. Microsoft used both its Windows event and Build keynote to demo a new GPT-4o-enhanced version of its Copilot assistant giving a first-time Minecraft player advice on how to play, right as a marauding zombie came into sight. Okay, maybe that specific use-case scenario isn’t world-changing—and I’m always suspicious about any breakthrough depicted in a canned video rather than performed in person on a stage. But at least the Minecraft demo displayed AI’s potential to figure out how to be helpful in real time. That could be transformative, and I’d like to try it myself in a variety of contexts. Which Microsoft says I’ll be able to do. Soon.

At least soon suggests a new AI feature is nearly finished and might show up within weeks or months. At Google I/O, I had loads of fun playing a Pictionary-like game with Google’s multi-model Astra AI assistant, which correctly identified my quick sketch of a dog wearing a top hat and eating an ice cream cone. During an I/O keynote video, Astra even helped a user find a pair of misplaced eyeglasses. Pretty amazing! But Google was notably noncommittal about when Astra might ship—or even whether it would offer the specific capabilities it demonstrated at I/O. The debut was less of a product announcement than a lab preview, helpfully reinforcing that yes, Google is working on something akin to ChatGPT Voice Mode. In other words: Don’t get too excited just yet.

Willful fuzziness about when new AI functionality might arrive is so pervasive that the rare instance of an announcement with a specific date attached comes as a pleasant surprise. Microsoft says the first Copilot+ PCs will be available on June 18, giving anyone who buys one the opportunity to judge for themselves whether they indeed usher in yet another new era of AI. I may or may not mark my calendar—but it’s nice to know I could.

  Be in the Know. Subscribe to our Newsletters.


Harry McCracken is the technology editor for Fast Company, based in San Francisco. In past lives, he was editor at large for Time magazine, founder and editor of Technologizer, and editor of PC World. More More

More Top Stories: