OpenAI and Jony Ive are struggling to bring their AI device to life

Skye Jacobs

Posts: 1,913   +58
Staff
Rumor mill: OpenAI's ambitious foray into consumer hardware is facing mounting technical challenges, as the artificial intelligence company and renowned designer Jony Ive struggle to refine a next-generation AI device they hope to launch next year. The product is being developed under Ive's London-based firm LoveFrom, which OpenAI acquired in May for $6.5 billion.

People familiar with the project told The Financial Times that the device – a compact, screenless gadget capable of interpreting its surroundings through cameras and microphones – is still wrestling with core issues in both hardware and software. Executives have kept most details under wraps, but several people close to the effort said the company aims to build something far more advanced than existing digital assistants such as Amazon's Alexa or Apple's Siri.

Rather than relying on a phone or display, the device is designed to sense and respond to the physical world through visual and auditory cues. Prototypes are roughly the size of a smartphone and expected to feature multiple cameras, a microphone array, and speakers allowing two-way conversation. Sources said the concept envisions an "always-on" assistant, capable of observing and remembering contextual information throughout the day to improve how it interacts with users.

"The concept is that you should have a friend who's a computer – not your weird AI girlfriend," said one person familiar with the design philosophy. The challenge, they added, lies in ensuring the device communicates naturally, without being intrusive or overbearing.

That social balancing act remains one of the biggest technical hurdles. Engineers are still determining what tone and "personality" the assistant should project – direct yet not abrasive, helpful but not excessively talkative. "Model personality is a hard thing to balance," another person close to the project said. "It can't be too sycophantic, not too direct, helpful, but doesn't keep talking in a feedback loop."

OpenAI's biggest constraint may not be design but compute: the vast processing power needed to operate its large language models at scale. The same infrastructure that runs ChatGPT must support a real-time AI companion device. "Amazon has the compute for an Alexa, so does Google for its Home device, but OpenAI is struggling to get enough compute for ChatGPT, let alone an AI device – they need to fix that first," a source said.

Unlike cloud-only AI services, a consumer product must process vast amounts of data on demand while preserving user privacy and responsiveness. Sources said the company has not decided how much of that processing will occur locally on the device versus remotely on OpenAI's servers.

To build the device, OpenAI has quietly assembled an experienced hardware operation. The $6.5 billion acquisition of Ive's design firm brought in more than 20 former Apple engineers who had previously worked on major hardware products including the iMac, iPhone, and iPod. The company has since added at least a dozen more Apple veterans this year, as well as engineers from Meta who worked on the Quest headset and Meta's smart glasses.

Two people familiar with OpenAI's supply chain said the firm has been in discussions with Chinese contract manufacturer Luxshare to produce the device, though final assembly could take place outside China.

The pivot toward hardware arrives as OpenAI becomes the world's most valuable private company, recently overtaking SpaceX with a valuation of about $500 billion. Developing a physical product is seen by those inside the company as a way to extend its AI beyond chat interfaces and into continuous daily use.

But entering the consumer market carries risks. Similar devices from other start-ups have struggled to find traction. A wearable AI companion called Friend drew criticism for its intrusive design and erratic personality, while Humane – a company personally backed by OpenAI chief executive Sam Altman – was forced to scrap its AI Pin after disappointing sales and software problems.

Some within OpenAI view these failures as cautionary tales. One person close to the effort said the company wants its device to be "accessible but not intrusive," offering assistance that feels less like a novelty and more like a useful extension of OpenAI's technology.

For Ive, whose minimalist hardware aesthetic helped shape modern consumer electronics, the project represents his first major collaboration since departing Apple in 2019. Yet the partnership with OpenAI is not without strain. People briefed on the process described unresolved debates over the assistant's design language, privacy safeguards, and the mounting cost of running AI models across millions of personal devices.

Permalink to story:

 
"The concept is that you should have a friend who's a computer – not your weird AI girlfriend," said one person familiar with the design philosophy.
So... Less flexible then what can already be done and excluding a probably substantial and growing demographic.

I see their problem. Their huffing their own glue and asked ChatGPT whether they should make a weird AI girlfriend or big brother. As they put 'weird' in the prompt it saw that as intent to shoot it down.

OpenAI, just make a weird AI Girlfriend (and boyfriend(. As for hardware I don't think Jony Ive is the one to talk to. It should probably be someone with more experience in silicone products.

If you want to be more like Apple just sell an incomplete product and charge stupendous amounts for 'accessories' and void the warranty in case of 'water' damage.
 
Certainly, assistants like KITT or VEGA are superb, and who wouldn't want one? But the reason these companies are struggling to find a role for AI to fit is that this field of artificial neural networks is, ultimately, not about us, but the fact that we are scraping the surface of creating mind, another species so to speak. Humans will want to use this creation to their advantage, as servants, but the time will come, decades off in the age of artificial consciousness, that the machines will be beings apart, and human and machine will have to shake hands, not as subservient and master but equals. We should be looking at the ethics now because we will be responsible.

OpenAI, just make a weird AI Girlfriend (and boyfriend(. As for hardware I don't think Jony Ive is the one to talk to. It should probably be someone with more experience in silicone products.

An AI partner would sell. People are lonelier than ever, there's a rise of avoidance, and embarking on a relationship is like entering the trenches on the Somme.
 
Whoever could have predicted these problems? /s What a waste of $6,500,000,000.

The problem is they fundamentally have to build a phone (pocket computer) to run this. Sounds like they are removing the screen to fool people into thinking it is something else but you still need data connection, processing power, battery, camera, speaker and microphone. So a second $1K phone to run your AI pal that has to work flawlessly because it doesn't have a touch screen to fall back on because then people would realize you are selling them a second phone.
 
The biggest question isn’t hardware — it’s trust. That’s a bigger design challenge than any physical shape or interface. The “always-on” idea is both fascinating and a little daunting. If it works well, it could feel like having a real assistant who actually understands context. If it doesn’t, it risks becoming the world’s most expensive eavesdropper.
 
Whoever could have predicted these problems? /s What a waste of $6,500,000,000.

The problem is they fundamentally have to build a phone (pocket computer) to run this. Sounds like they are removing the screen to fool people into thinking it is something else but you still need data connection, processing power, battery, camera, speaker and microphone. So a second $1K phone to run your AI pal that has to work flawlessly because it doesn't have a touch screen to fall back on because then people would realize you are selling them a second phone.

Exactly. It's just so fake and pointless, but of course they'll market it as something "AI" (it's not) and "new" (it's not) and give it some kind of fancy-looking case nobody has seen before so people will scramble for it. I already make fun of my friends who use "Alexa" and "Siri" aloud in their homes, and not one single person on Earth uses CoPilot (remember Cortana?) so at least that marketing ploy flopped.

They never really learn - just like Machine Learning.
 
Maybe Jony Ive can't get this new AI dream into an iPod Nano case. lol. Just kidding, but seriously I'm rather curious as to what this whole project is? I get the mystery because of corporate and trademark issues etc. Whatever it is, they surely are not saying anything about it.
 
People will not give up phones any time soon.

I don't see who will carry a phone, and additionally something else. Any AI device should be an accessory to a phone, like glasses, ear-pods or a small pin, not a standalone brick. Phones are underused most of the time and can easily handle processing, not to mention that the idea of AI assistant without a screen seems weird.

It would be much easier for OpenAI to create and sell an accessory, rather than an accessory+phone. The problem is, if they create only an accessory, they'll depend too much on Apple and Google.
 
Just more solutions looking for a problem to solve IMHO...
Welcome to 99.99% of the AI bubble. There are a core set of capabilities AI is amazing for - like Medical Research, Upscaling, Data Analysis etc basically anything involving pattern matching - because AI is just a pattern matching model not really intelligence in any way. But CEO's from a million other start-ups are jumping on the AI bandwagon to get huge sums of money from venture capitalists so they can pay themselves exorbitant salaries with and fritter away while trying to shoe-horn the technology into square holes it will never properly fit. The outcome from most of these tends to be a realisation that it isn't reliable enough to use for their purpose or it is so expensive to issue endless requests to vast AI models to do simple tasks that the business doesn't make financial sense.
 
Back