Science and Technology Innovation Board Daily, May 12 (editor Song Ziqiao) OpenAI is determined to launch a live broadcast and launch a new product the day before Google's Imax O conference to "demonstrate magical ChatGPT and GPT-4 updates."
What is this mysterious new product?Onlinepokerroulette? Speculation about GPT-5 and search engines has been personally overturned by OpenAI CEO Altman.
According to the latest reports, the AI assistant built into the phone may be the upcoming product of OpenAI.
Technology media The Information quoted people familiar with the matter as saying that OpenAI plans to launch a multimodal AI model that has visual and auditory functions, can not only talk to you, but also identify objects, and has better logical reasoning ability than current chatbots. OpenAI has shown this model to some customers.
OpenAI has developed models that can transcribe audio and text-to-speech, and it is reported that the new model is equivalent to a combination of these models, but is more accurate and responsive. The new model can help AI assistants distinguish mood and better understand semantics, and "in theory", it can help students learn math or translate real-world gestures.
However, although the new model can surpass GPT-4 Turbo in "answering certain types of questions", it still has hallucination problems.
Ananay Arora, a developer, said that OpenAI may launch a service with a built-in ChatGPT function on the phone to make calls. Arora posted a screenshot of the above call-related code on the social platform and found evidence that OpenAI has a server configured for real-time audio and video communications.
Using artificial intelligence to make phone calls can save users waiting time, and this service can be regarded as one of the functions of AI assistants.
AI Assistant is also a feature that Google has been developing. It is reported that Google's Pixel 9 series of phones will have a new dedicated AI assistant "Pixie" that can view items through the device's camera and perform operations such as indicating where to buy or providing instructions for the use of items.
Altman had previously revealed in an exclusive interview with Marc Benioff, chief executive of Salesforce, that his favorite AI movie was "Her," the story of a man falling in love with his AI virtual assistant. "the idea of a conversational language interface is incredibly prescient."
The Information reported that Altman hopes to eventually develop a virtual assistant that can respond quickly like the AI assistant in the movie, and use this technology to support existing voice assistants such as Apple's Siri.
It is worth noting that Apple is about to reach an agreement with OpenAI to introduce the latter technology into the new generation of iOS operating system, according to people familiar with the matter. The two sides have been hammering out the terms of an agreement to use ChatGPT functionality in Apple's next-generation iPhone operating system, iOS 18.
▍ 's new model runs on the cloud and is expected to be included in the free version of ChatGPT in the future.
OpenAI believes that AI assistants with visual and auditory functions have the potential to bring about change like smartphones. It can observe users' environmental information, provide advice, and potentially use it, such as acting as a tutor, translating traffic signs, repairing cars, and so on.
But similar technologies currently require too much hardware to run on personal devices. Media analysis pointed out that the operation of the new model depends on the cloud and requires an Internet connection to work. It can take months or even years for complex AI conversations with visual and auditory functions to be small enough to run on personal devices such as mobile phones.
It's not clear when OpenAI will offer these new features to paying customers, but according to people who have tried the voice assistant, OpenAI's ultimate plan is to include them in the free version of ChatGPT, which aims to be cheaper to run than its state-of-the-art model, GPT-4 Turbo.
OpenAI did not respond to the speculation.
What will OpenAI eventually launch? The answer will be revealed next week, and OpenAI has announced that it will broadcast live on its website at 10:00 Pacific time on May 13, when it will demonstrate some ChatGPT and GPT-4 updates.