5 mistakes Sam Altman and Jony Ive need to avoid to stop their ChatGPT AI device going the way of the Rabbit R1 and Humane AI Pin
It’s not often that two of the biggest names in tech get together to “completely reimagine what it means to use a computer”, but that’s what’s happened when Sam Altman’s OpenAI bought Jony Ive’s io design company recently and announced that they would be creating some new AI products together.
Jony Ive is perhaps most famous for designing the iPhone, and Sam Altman is the driving force behind ChatGPT, so my mind immediately hopped to the idea of the world’s first ChatGPT-powered smartphone, but it turns out that Ive and Altman have got something else in mind. Something that we might have seen before.
It was all revealed in a nine-minute video that OpenAI released yesterday called Sam and Jony introduce io:
The video starts with Altman saying, “I think we have the opportunity here to kind of completely reimagine what it means to use a computer”.
According to Ive, io is “formed with a mission of creating a family of devices that would let people use AI to create all sorts of wonderful things”. The video goes on to mention a new prototype of Ive’s that Altman has been testing.
Altman calls it “the coolest piece of technology that the world has ever seen”, but frustratingly doesn’t go into any details about it, although clues are littered throughout the video.
For example, in one segment, he talks about how cumbersome the process of using a laptop is because it has a screen.
It’s hard not to get swept up in the reality distortion field that surrounds these tech gurus, but creating an AI computer with no screen essentially means something small that you wear or carry around with you and interact with using your voice, and that concept has so far scuppered more than one tech startup.
Of course, I’m referring to the now legendary tech failures that are the Humane AI Pin and the Rabbit R1.
Beam me up Scotty
The Humane AI Pin was basically like the classic Star Trek combadge. It clipped onto your clothing and stayed there in standby mode until you engaged with it. It was ahead of its time in many respects – you could ask it questions (like an AI-powered Siri or Alexa) and get it to play music. It also had a camera, so it could see what you were looking at and answer questions about it. It could also shoot out a laser onto your palm that brought up a menu system that you could interact with using hand gestures.
The Rabbit R1 was similar, except it was designed to fit your pocket, not be worn on your chest.
They were both good ideas, and ahead of their time, but good ideas don’t automatically turn into good products. They were each flawed in some fundamental ways.
Here are five classic mistakes that people have already made with AI devices that Ive and Altman will need to avoid:
1. Sometimes you do need a screen
The Humane AI Pin was built around the idea of replacing your smartphone, but that was never going to work.
In theory, combining the power of sophisticated AI and the entire Internet, in a way that means you could still be completely present in the moment and not bogged down by intrusive and addictive phone applications, was a good idea. In practice, it just didn’t work.
We live in a world where everybody has already got a smartphone, so if you want to interact with other people, who are all using screens and sharing visual information, then you are going to have to use a screen at some point.
Whether that means that the io device can co-opt the existing screens in your life, or it can attach to a screen in some way, I don’t know, but remembering that screens are still essential for some things will be key to its success.
2. It can’t be too expensive
This is going to be key to the device’s adoption. A major criticism of the AI devices released so far is that they were too expensive. We already have an expensive smartphone and a data plan to go with it, please don’t expect us to be able to afford an equally priced second device and a second data plan each month.
Jony Ive gave us the iPhone and the MacBook Pro – devices that defined technology for a generation, but those products have always had a premium price tag.
The man may be a genius, but if this product is to succeed, it needs a price point that makes sense, not one that alienates its target audience.
3. The voice interaction has to be flawless
AI voice modes, particularly ChatGPT’s Advanced Voice mode and Gemini Live, have come a long way in recent months. They feel a lot more natural to use, as if you are having a real conversation with a friend.
Delivering that experience with the new io device will be key to its success. If conversation with the AI is stilted in any way or if there are long pauses after you tell it to do something – a problem that has plagued previous AI devices – then it’s not going to take off.
The whole experience of using it will need to be as frictionless as possible.
4. It can’t hallucinate at all
AI has a tendency to make stuff up, which is known as hallucinating, and it’s a problem that seems to be getting worse the smarter the AI gets.
According to its own reports, ChatGPT’s new o3 model incorporated hallucinations in a third of a benchmark test involving public figures, which is double the error rate of the earlier o1 model. The compact o4-mini model performed even more badly, hallucinating on 48% of similar tasks.
One theory that explains why AI is getting worse at hallucinating is that the deeper the reasoning method goes, the greater the chance it has to mess up. The older, simpler models stuck to higher-confidence predictions, unlike today’s complex models. When there are multiple paths to consider, the model has a greater chance of improvising and getting things wrong.
Whatever the issue, the Humane AI Pin, which could use any AI model, certainly hallucinated, and there needs to be no trace of the problem in any product released by io if it wants to be trusted.
5. It needs to work without an internet connection
This will be a tough one to implement for sure, but requiring a second data plan just for the device was one of the major problems with the Humane AI Pin. Without access to the Internet, it was useless. Perhaps the new io devices will be able to piggyback off your phone’s Internet connection, but ideally, they’d also be capable of some on-device processing, meaning that they could continue to function even in places where there’s no data coverage.
Back on May 7, Apple Exec Eddy Cue let slip that “you may not need an iPhone 10 years from now” whilst giving testimony at the Google Search antitrust remedies trial.
Clearly, the time is ripe for some kind of AI device, and the possibilities for reimagining a human and computer interaction are endless, but as we’ve seen, others have gone down this road before, and so far, it’s not ended well. Let’s hope that whatever io releases in 2026 (its promised delivery date for its family of AI products) can buck this trend because nobody wants to see another doomed AI device hit the market.