- The new AI model powering ChatGPT is doing its most to become your new best friend.
- GPT-4o, unveiled by OpenAI on Monday, can have conversations with you in a very human tone.
Whether you like it or not, Sam Altman just let you know you're getting an AI best pal.
At a livestreamed event on Monday, the OpenAI chief sent out his top technology lieutenant Mira Murati to unveil GPT-4o, a powerful new AI model that looks set to turn ChatGPT into a pocketable buddy.
The new model, which Altman hyped up as feeling like "magic" ahead of the event, was presented as a realization of the AI voiced by Scarlett Johansson in Spike Jonze's 2013 movie "Her," as demo after demo showed it interacting with you just as a friend would.
A simple "hey, how's it going?" can now trigger ChatGPT — newly powered by GPT-4o — to ask how you're doing in return, elicit comments about the clothes you're wearing, and offer opinions on any number of other things it fancies.
"I see you're rocking an OpenAI hoodie, nice choice," ChatGPT told a user in one demo. "What's up with that ceiling though? Are you in a cool, industry-style office or something?"
In another demo, an OpenAI engineer managed to get a helping hand from ChatGPT by getting it to tell a bedtime story about robots and love to a colleague named Barrett — after being told he'd been struggling to sleep.
"I got you covered!" the AI bot said, before delivering a story about a robot named Bite in dramatic, singing, and robotic voices, depending on the request from the user.
It also proved to be a quick and nimble by translating Murati's Italian in live conversation.
Much of this has been enabled by something OpenAI's previous models have lacked: an ability to "reason across audio, vision, and text in real time."
That it has a voice similar to Joaquin Phoenix's AI pal Samantha from "Her" goes a long way towards elevating ChatGPT engagements towards more human-like experiences too.
Taken together, these demos offered the strongest sign yet of Altman's ambitions for AI chatbots: turning them into entities that users begin to engage with in a more continuous way, like they might when hanging out with a friend for the day.
Ethan Mollick, an associate professor studying AI at Wharton University, has seen signs of this coming for some time.
After the event, he wrote that GPT-4o marked an "important shift" in which newer AI features rolled out by the likes of OpenAI seem targeted at "making AI more present and more naturally connected to human systems and processes."
"If an AI that seems to reason like a human being can see and interact and plan like a human being, then it can have influence in the human world. This is where AI labs are leading us: to a near future of AI as coworker, friend, and ubiquitous presence," Mollick wrote on Substack.
OpenAI is not the first company to try to position its AI as a buddy. Inflection AI, for instance, revealed its bot Pi a year ago as "a kind and supportive companion that's on your side."
Though Inflection struggled to gain traction for the idea (the company lost its CEO Mustafa Suleyman and a bunch of engineers to Microsoft in March), OpenAI's reveal this week has shown that technical advances can start to make an AI bot feel like a more human partner.
It's something its rivals will pay close attention to, with the possibility of a new battle emerging between companies trying to release the most human, friendly AI bot they can. A recent report suggested Apple is racing to improve Siri with generative AI features.
What kind of future this all might lead us to, however, remains to be seen. As Mollick put it: "I don't think anyone, including OpenAI, has a full sense of all of the implications of this shift, and what it will mean for all of us."