OpenAI unveiled GPT-4o on Monday, showcasing the real-time audio features that make it seem like Spike Jonzeās film Her is now becoming a reality. Just like in the movie, OpenAI gave ChatGPT a dynamic, friendly, and arguably flirty, voice that truly sounds just like a human. Iām going to make an early and somewhat obvious prediction: people are going to fall in love with GPT-4o, just like in the movie.
No, I donāt mean to fall in love like you might love Google Search. You love to use Google Search, but I think people are going to literally fall in love with GPT-4o or whatever it is they project onto the lifeless bot.
AI girlfriends are already proving to be a concerningly large use case for generative AI technology. When OpenAI released the GPT Store, AI girlfriends quickly flooded the zone. They offer companionship, without the real weight of a true human partner. Now that ChatGPT can supposedly talk in real-time with emotional intonations, I think OpenAI just blew the fucking lid off this thing, but maybe that was the point.
It shouldnāt be understated how impressive this technology is. OpenAIās demos supposedly show that robots can now talk with humans in real-time, in a conversation thatās almost indistinguishable from a phone call with a real person. If the final release runs like Mondayās demo, OpenAI seems to have connected the dots for many people on how much AI will change our world.
OK, GPT-4o pretty darn impressive (and available now in Playground)
Initial thoughts are not surprising after preview: VERY fast, clearly smarter than GPT-4 (though not GPT-5 smarter), better at vision. Big practical step forward, though not a giant leap in underlying abilities. pic.twitter.com/wWu0LyYa7n
— Ethan Mollick (@emollick) May 13, 2024
The OpenAI staff noted how this technology was just like the movie Her. Sam Altman tweeted the word āherā right after the launch. Another OpenAI staff member who works on āAudio AGI researchā has a picture from the movie in his bio. Make no mistake, this was the goal. As often happens in tech, however, hot companies like OpenAI move fast and amaze us with technology without considering the repercussions.
As we saw in the movie Her, Joaquin Phoenixās character slowly fell in love with an AI companion voiced by Scarlett Johansson. Itās impossible to not notice the similarities between Johanssonās voice and the one that OpenAI debuted on Monday morning. At one point, GPT-4o laughed at a joke an OpenAI staffer made. At another, GPT-4o said, āThatās so sweet of you.ā Iām willing to bet a lot of people out there donāt have someone in their life who says either of those things. Iād also bet a lot of them would jump at the chance to have at least a single voice that does soāartificial or not.
Companionship is a somewhat obvious use case for a chatbot. But real-time voice conversations offer more intimacy than text on a screen. The near-zero latency makes talking feel natural, like a long-distance relationship whenever you want it.
The emotional nature of GPT-4o makes the AI chatbot much more personal than Siri ever was. Siri felt like talking to a robot, but these demos made clear that GPT-4o has artificial emotional intelligence, able to identify your feelings and match your energy. This makes GPT-4o feel like a true companion, adding a touch of humanity to your smartphoneās operating system.
Weāve already seen a community meltdown when Replika issued an update to its much more primitive chatbot in 2023 that many users felt was like losing a friend.
But thereās one human emotion GPT-4o lacks: frustration. The OpenAI staff showed how GPT-4o doesnāt mind being interrupted, a very real complaint actual women have with male partners. This shows how some men who are incapable of holding a meaningful relationship with a woman might turn to an AI chatbot like this. This could offer lonely individuals a meaningful connection. But it also could make it that much easier for people to never attempt the growth necessary to try at the real thing. GPT-4o could make our devices even more tantalizing than the human connections many of us already lack, just as a matter of convenience.
Iād be remiss to not express concern about a technology like this. Humans report increasing levels of loneliness in the 21st century. One in three Americans feels lonely every week. While companionship may not be the main thing OpenAI is selling with GPT-4o, that might be what people grab onto most.
This is not the first time our devices have offered companionship, while in reality, further isolating us from each other. Social media capitalized on the innate human desire for connection, but that turned out to be a great illusion. Low-stakes human connection from behind the screen often doesnāt really work out and just as often pits us against each other. AI companions like GPT-4o take this phenomenon one step further, completely removing the human from the equation.
Thereās been a lot of fearmongering around generative AI, but this one has flown under the radar. OpenAIās demo marked the arrival of a newfound technology, and potentially, a new societal problem weāll be grappling with for decades. The worst side effect Iāve seen from technology in my lifetime is its ability to isolate humans from each other. GPT-4o just made that isolation that much easier.
