Where do AI gadgets go from here?

This spring, the hype revolved around two AI-powered gadgets: the Human AI Pin And Rabbit R1. Both promised AI automation and seamless conversation with an always-present, always-helpful AI assistant.

They failed. Prominent technical critic Brownlee Brands called the Humane AI Pin “the worst product I’ve ever reviewed,” while the Rabbit R1 received the somewhat kinder verdict of “barely reviewable.”

“I don’t think it will be more than six months before we see real (AI) applications running on PCs, or even mobile devices.” —Davec Chenna, AMD

Dr. John Pagonissenior UX researcher at Zanshin Labs, observed that any new consumer device must prove that it is better than what is already available, a task that both AI gadgets failed to do. “What problem do (these devices) solve? What needs do they cover? It’s not easy.

So it’s over, right?

Not quite. Although Humane and Rabbit failed, the solutions to the problems that stymied these newcomers are within reach and could change consumer technology forever.

ChatGPT, are you there? Good morning? Good morning…?

Today’s best AI extended language models (LLMs) face a common enemy. Latency. People expect feedback when they type or speak, but the best LLMs reside in data centers, which can cause delays. This is at the heart of Humane and Rabbit’s misfortunes. Reviewers complained that the gadgets were slow to respond and useless when Internet access was unreliable or unavailable.

However, there is a solution: place the LLM on the device. I pointed out this possibility for IEEE Spectrum in December, and a lot has happened since then. The llama of Meta 3, Microsoft Phi 3And Apple OpenLEM– all announced in April 2024 – have brought significant gains in the quality of small AI models. Chipmakers like Apple, Intel, AMD and Qualcomm are also tackling the problem by improving the performance of AI coprocessors in laptops, tablets and smartphones.

“All these apps, all these different notifications. It’s too much. It’s exhausting. Lots of research shows that this shouldn’t be the way we interact with technology. —Patricia Reiners, UX designer

Dwith Chennaan engineer at AMD specializing in AI inference, said these improvements make LLMs possible without the cloud. The Humane AI Pin and Rabbit R1 were simply released too early to take advantage of these advances.

Apple’s M4, which debuted in the new iPad Pro, has, according to the company, “Apple’s fastest neural engine ever.”Apple

“There’s a lot of effort going into trying to compress the models (in broad terms) to run them on devices like PCs and mobile phones,” Chenna said. “I don’t think it will be more than six months before we see real (AI) applications running on PCs, or even mobile devices.”

Bringing LLMs into mainstream technology will also address another key issue. Confidentiality.

Rabbit’s R1 uses a Large Action Model (LAM) to automate applications and services, but some critics have expressed unease with the idea. LAM requires personal information, including usernames and passwords, to act on a user’s behalf. Promised Rabbit manage information securelybut because the AI ​​model it uses is hosted in the cloud, the data is inevitably sent outside the device.

“I think one of the main concerns is privacy and security. Not everyone is comfortable sharing their information with the cloud,” Chenna said. Pagonis agrees and points out that the biggest tech companies are already maneuvering to solve the problem. “At Google I/O they were talking about running Gemini Nano on device for better privacy. And it’s a strategy that Apple will follow, I’m sure.

Design trips up AI gadgets

Smaller, faster LLMs that run on the device might solve the latency problem, but they won’t instantly redeem Humane and Rabbit’s gadgets. Both made serious design errors that hinder the promised ease of use.

“I think the technology is absolutely fascinating and also… super revolutionary,” said Patricia Reinersindependent UX designer and facilitator Podcast on the future of UX. “But it has to work. Basic usability issues shouldn’t be the case.

Reiners explained that Humane and Rabbit’s misfortunes could be avoided if they took a step back from their AI ambitions to think about how people use technology in the real world. The Humane AI Pin can overheat with frequent use and relies on a projector to display information, which is problematic when using the Pin outdoors. Rabbit, meanwhile, disables the R1’s touchscreen in some menus, but not others, confusing users.

The Humane AI pin uses a projector to display information. The projected images would also be difficult to see outdoors.Human

“I think it’s extremely important for people who read your article,” Reiners said. “Test devices or products from the beginning, from the first idea. Start with prototypes.

Reiners and Pagonis had different opinions on the fate of Human and Rabbit. Reiners doubts Humane’s design issues can be fixed, but said the Rabbit R1’s “half-baked” features could be fixed with updates. Pagonis was more skeptical. “(Humane and Rabbit) failed, in my opinion, in the fundamental exercise of product discovery, which is figuring out how useful a product is and then making it easy,” he said.

Could AI kill the app?

But they agree on one thing: the failure of the first AI gadgets leaves the arena wide open for Apple, Google and Microsoft.

“Apple and Google are not sleeping,” Reiners said. “This is why the Rabbit R1 and Humane AI Pin were in such a rush to ship. They know that the big players work there.

Pagonis went further and predicted a clean slate for big tech. “Who will win? Companies that own your data, like Google. This controls the user experience, like Apple. And have a relationship with you, like Apple and Google and, of course, Microsoft.

This may seem disappointing. This implies that future AI-enabled devices will maintain the status quo: they will look, feel and work like the smartphones we are used to.

Google explained how it is introducing LLMs to its smartphones during Google I/O 2024.CNET

But Reiners thinks that’s not the end of the story. AI may not reinvent what tomorrow’s consumer technology will look like, but it could help reinvent the software we use on our computers, tablets and smartphones.

“When you open your phone, so much happens. All these apps, all these different notifications. It’s too much. It’s exhausting,” Reiners said. “A lot of research shows that this shouldn’t be the way we interact with technology.”

Reiners believes companies like Apple and Google will try to deliver on the promises made by Humane and Rabbit with streamlined AI operating systems that predict user needs and automate common tasks. She notes that smartphones would prove easier to use if they gave users fewer options after unlocking the device. Phones can even replace apps with automations controlled by an on-device AI agent.

“The user doesn’t really need apps,” Reiners said. “They have a goal, what they want to do and want to achieve it. So, as designers and people working in technology, we need to rethink how we interact with users.