More products, more problems.
Behind the scenes, OpenAI is working on a new AI device in collaboration with the legendary former Apple designer Jony Ive. But according to the Financial Times, it’s running into some major technical issues that could delay its release — along with some more philosophical ones, too.
The goal, per the reporting, is to make a palm-sized device without a digital display that constantly monitors its surroundings for audio and visual cues, responding to user requests where appropriate. A source described it as “like Siri, but better.”
But one of the quandaries that OpenAI and Ives are running into? Not making the AI too personable and sycophantic — which is telling, since finding that balance is a problem besetting the chatbot industry at large.
“The concept is that you should have a friend who’s a computer who isn’t your weird AI girlfriend,” the source told the FT. It should be “accessible but not intrusive.”
OpenAI acquired Ive’s startup io for $6.5 billion in May, but the details of their arrangement has largely been kept under wraps. That month, CEO Sam Altman told his staff that he planned to ship 100 million devices designed with Ives, which he described as AI “companions.” This appears to be the same device we’re hearing about now.
A key part of the design philosophy is featuring zero screens. To compensate, the small device will rely on a camera, microphone, and a speaker to interact with users. It may even use multiple cameras, according to one FT source.
Instead of being triggered by a phrase like Apple’s “Hey Siri,” the device will be “always on,” constantly gathering data so it can build its “memory,” something that other wearable AI tech has relied on, and will undoubtedly raise massive privacy concerns if it ever hits shelves.
But it sounds like OpenAI and Ive want their gadget to be more than just a wearable. It’s also being designed to sit on a desk or table, and will be roughly the size of a smartphone. Not only is it supposed to be a better Siri, as a source described, but an improvement on smart speakers like Amazon’s Alexa or Google’s Home devices.
All this is easier said than done. For one, OpenAI already struggles to rein in ChatGPT’s engaging disposition without angering its own fans. Now it needs to somehow bridge this gap in a new and lighter-weight model. “Model personality is a hard thing to balance,” a source close to the project told the FT. “It can’t be too sycophantic, not too direct, helpful, but doesn’t keep talking in a feedback loop.”
And then there’s the hardware side of the equation.
“Compute is another huge factor for the delay,” a source close to Ive told the FT. “Amazon has the compute for an Alexa, so does Google, but OpenAI is struggling to get enough compute for ChatGPT, let alone an AI device — they need to fix that first.”
Even if those get ironed out, will there be demand to justify the billions OpenAI has invested to make this collaboration happen? ChatGPT boasts over 700 million weekly active users, the company claims, so technically it isn’t unthinkable that OpenAI could ship 100 million little AI gadgets. But as of April, it only had 20 million paying subscribers, which suggests that the vast majority of its userbase aren’t willing to shell out money for OpenAI’s products. Perhaps a physical device might be more alluring, but it’s a big gamble.
More on OpenAI: OpenAI’s Sora 2 Is Generating Video of SpongeBob Cooking Meth, Highlighting Copyright Concerns