OpenAI’s GPT-4o, a new version of the company’s artificial intelligence chatbot, is a significant step forward from its predecessors. Demonstrations show the AI interacting with users in ways that seem startlingly human.
Unconfirmed reports indicate Apple and OpenAI are negotiating a deal that would build some version of GPT into iOS and macOS.
OpenAI GPT-4o takes AI chatbots to new heights
In 2022, OpenAI’s ChatGPT wowed the world by demonstrating the possibilities of generative AI based on a large-language model. It’s progressed enormously since then.
“GPT-4o (‘o’ for ‘omni’) is a step towards much more natural human-computer interaction — it accepts as input any combination of text, audio, and image and generates any combination of text, audio, and image outputs,” OpenAI said Monday.
To show off the potential, the company demonstrated that the new system is extremely conversational. It casually chats with an employee about the announcement:
In addition, OpenAI released a video demo of GPT-4o explaining to a student how to do a geometry problem:
Other demos show off the system’s ability to read and understand text using an iPhone’s camera, do real-time English/Spanish translations, and more.
Could be the next Siri
Early versions of ChatGPT showed the limitations of Apple’s Siri, and GPT-4o makes an even starker contrast. Fortunately, leaks indicate Apple long ago realized its existing voice assistant is utterly unable to compete with an AI powered by a large-language model. It’s reportedly in talks with OpenAI to use the company’s technology in iOS 18 and macOS 15.
There’s little room for doubt Apple devices can handle the artificial intelligence software — OpenAI’s demonstrations Monday were done using iPhones, though the company didn’t draw attention to this fact.
It’s possible GPT-4o will simply replace Siri, but it’s more likely to operate in conjunction with Apple’s assistant. Previous leaks indicated that Apple wants to keep AI functions running on users’ iPhone or Mac as much as possible for privacy reasons, which would preclude exclusively using GPT-4o. An on-device AI system could handle basic Siri commands, while more complex requests could go to data centers with servers running on Apple’s own processors.
We’ll find out in less than a month at WWDC24. Apple’s annual developer conference kicks off June 10, with the new AI capabilities of iOS 18 and macOS 15 expected to be high on the agenda.
