OpenAI should make a phone

The writer is co-founder & CEO of Mill and co-founder of smart device company Nest, purchased by Google

The first wave of AI hardware released into the world has flopped for one simple reason: the products were built on hype, not function. 

Take Humane’s AI Pin, for example, which promised to replace the smartphone. In reality, the wearable gadget failed to integrate with the products that people use their phones for, such as email, and demanded users learn awkward new gestures. Plus, nobody wanted to talk into their shirt — in public or otherwise. 

Similarly, Rabbit’s R1 was marketed as a pocket-sized AI companion but launched before it had any clear reason to exist. And the new AI companion Friend, a $129 wearable necklace described as looking like a Life Alert button, also fails to answer the question: “What am I supposed to do with this?”

All three products have been over-promoted, under-built and detached from customer problems. They focused too much on packaging, industrial design and publicity. This is the inverse of how Apple approached the first iPhone. I worked alongside Steve Jobs and his engineering team on that launch in 2007 and saw first-hand the ways in which it was designed to solve obvious problems, such as clunky keyboards and painful web browsing on tiny screens. It passed what Google co-founder Larry Page once called the “toothbrush test”: it solved problems that real people encounter every single day. 

So far, AI hardware has failed to come anywhere close to this test. The products created have submerged swiftly and publicly into the flunk tank.

Despite these failures, I believe the trajectory of generative AI is still linked to hardware. In order for the technology to live up to its promise, it must be part of the physical world around us. That means it must be integrated into our phones. 

The hardware graveyard of the past decade is littered with products that promised to replace phones yet never made it past the first few shipments. The smartphone is an enduring device because the interface is visual and easy to use — it’s been embedded in people’s psyche (and pockets). 

So far, OpenAI’s $6.4bn bet on former Apple designer Jony Ive’s start-up io earlier this year represents the biggest AI hardware gamble to date and a chance to define how the technology integrates into human lives at scale. But in order to be successful, Ive and OpenAI will need to go beyond building an AI-powered smart speaker or accessory — rumoured to be where the team are focusing their time and effort — and focus on a phone. 

Ive’s team is world-class. I know because I worked alongside most of them in Apple’s engineering labs. At the moment, however, they’re light on the software talent that was crucial to the iPhone’s longevity — something Ive and OpenAI chief executive Sam Altman need to rectify. Hardware only succeeds if it does something people need on a daily, or near-daily, basis — easier, faster, better and more fun than before. Engineering cleverness and media hype are not enough. 

There is a final hurdle. To build an AI-powered phone outside of Apple’s walls requires Google’s infrastructure. Android is the only commercially available open-source mobile operating system. Ive’s background at Apple and OpenAI’s rivalry with Google may make this difficult to swallow but it would be the fastest, best way for Ive to make a return on OpenAI’s investment.

Producing an AI phone is the tallest order in tech since 2007. But in order for AI hardware to succeed and deliver on the promise of inserting the technology in everyday life for billions of people, it must fit into life as we know it.

In other words, instead of taking months to build something that fails to complement or compete with the smartphone, OpenAI needs to engineer its next iteration.