OpenAI 与 Meta 押注 AI 智能体先行、硬件随后

While AI has made plenty of inroads on people’s phones and computers, it’s largely failed in dedicated devices. But over the next year, two major AI companies, Meta and OpenAI, will attempt to change that. And they’re apparently betting on a similar path: testing appetites for physical hardware with their cutesy software agents.

“Hardware is hard” is a tech industry mantra. Dedicated AI devices like the Friend and Humane AI Pin have inspired mainly frustration and backlash, which may only worsen as public anger toward AI grows. Meta and OpenAI, however, are both pushing ahead. OpenAI is working with famed ex-Apple designer Jony Ive and plans to ship no earlier than February 2027, per a public filing . Meta aims to launch a pendant-like device sooner, before this year’s holiday shopping season.

While OpenAI’s hardware didn’t make an appearance at this year’s DevDay, the company debuted a major new AI agent platform called Dots, complete with personalizable colored blobs with eyes. During a Q&A with reporters after the keynote, when asked whether Dots would make their way into a future physical hardware device, OpenAI CEO Sam Altman replied that that “seems like a very reasonable thing to assume we may do someday,” declining to offer further details.

Meta, meanwhile, is more directly tying its hardware to the recently released Muse agent. Compared by CEO Mark Zuckerberg to a keychain, the device is called the Muse Charm, and it bears a screen featuring its cuddly-looking Muse AI agent, giving it the air of a Tamagotchi. “We’ve packed a lot of technology into this little guy,” Zuckerberg said onstage . “So, if you’re not wearing glasses, this is going to be by far the fastest way to talk to your Muse and to show it what’s going on around you.”

Both OpenAI and Meta have likely learned from other companies’ failed hardware rollouts, and their resulting strategy seems to be soft-launching AI agents before hard-launching the hardware. Their solution for hacking the hardware problem is to prime the market by getting people used to AI agents, hoping people will spread the word that “yes, they really work now.” Once people find value in — and even get attached to — the cutesy agents, they might want them in a more physical form. Other companies, like Friend, have taken similar software-first strategies, but Friend was positioned mainly as a companion chatbot, not a productivity tool.

When asked during the DevDay press Q&A whether Dots will likely become the primary interface for interacting with ChatGPT, Thibault Sottiaux, OpenAI’s head of core products, said the company would be “evolving ChatGPT in place to be this incredible personal assistant very soon” and that he believes that as people are able to reach their AI agents via email, text, or phone call, the point of interaction is “all going to blur at some point.” It’ll be a “natural part of your life,” Sottiaux said.

Altman is banking on the fact that Dots will be more useful for long-term tasks than previous iterations of agents on the market have been. “This feels like an always-on helper, collaborator … that you can give very large pieces of work to and it gets to know you and helps proactively over time,” he said during the Tuesday press Q&A.

Both companies’ main bet is that, after years of less-than-impressive AI agents, they can finally deliver a useful everyday tool that will break through to a broader public market. But after Muse’s concerning exploits in Facebook Marketplace, and a Dots voice demo that went awry at DevDay, there’s apparently still some work to do.