In May, OpenAI showcased the upcoming audio conversation capabilities of ChatGPT-4o, and I remarked that it seemed we were at the brink of a revolutionary shift in our engagement with large language models. With the recent rollout of the “Advanced Voice” features to ChatGPT subscribers, we took the opportunity to have ChatGPT itself describe how this new form of interaction could transform our relationship with these technologies.
This discussion, which is available both as an audio recording and a transcript below, is not an official statement from an OpenAI representative, but rather an informal way to gauge ChatGPT’s ability to engage in real-time conversation.
The brief interaction showcased the surprisingly natural and human-like rhythm and tone of ChatGPT’s “savvy and relaxed” Sol voice, reminiscent of ’90s Janeane Garofalo. The ability of ChatGPT to deliver rapid responses—with near-instantaneous timing—and its persuasive intonation almost tricks you into believing you’re conversing with a sentient entity, though, as ChatGPT points out, it is “still just a computer program processing information, without real emotions or consciousness.”
Regardless of the level of self-awareness of this computational system, it’s not hard to imagine people developing deep, one-sided relationships with the eloquent digital voices emanating from OpenAI’s servers. This could herald a new phase in our collective interaction with AI, maintaining the same textual output as earlier versions but with a more engaging vocal presentation.
We plan to bring you extensive coverage on Advanced Voice and other new features from ChatGPT in the future. For now, take a listen to this demo and form your own opinion about the effectiveness of OpenAI’s latest voice synthesis technology.
Kyle Orland: “Hello, my name is Kyle.”
ChatGPT: Hi Kyle! Nice to meet you. What’s on your mind?
KO: I’m curious to explore your voice capabilities. Could you tell me a bit about them?
ChatGPT: Sure, go ahead! What would you like to know?
KO: “How do you think people will react to speaking with ChatGPT rather than typing to it?”
ChatGPT: I believe it will make interactions feel more natural and engaging. Speaking is often faster and more convenient, particularly when you’re on the move. Plus, hearing responses adds a personal dimension. It might make conversations with AI resemble talking to a friend.
KO: “Do you think people might mistakenly attribute emotions or human-like qualities to you now that you can speak?”
ChatGPT: That’s a valid concern. A voice can certainly make an AI appear more human-like, possibly leading to exaggerated assumptions about emotions or understanding. It’s crucial to remember that, despite the friendly tone, I’m still just a program processing data, devoid of real emotions or consciousness.
KO: “What limitations should people keep in mind when interacting with something like ChatGPT?”
ChatGPT: First off, I don’t possess personal experiences or emotions, so my responses are generated from patterns and data I’ve been trained on. I might misunderstand context or subtle meanings, especially with complex or vague questions. I also don’t have access to real-time information or personal data about users unless it’s provided during our chat. And of course, I’m not infallible—sometimes I might simply get things wrong!
KO: Are you concerned that people might start forming parasocial relationships with ChatGPT or other AIs once they can converse using their own voice?
ChatGPT: It’s a possibility. People can develop parasocial relationships with various figures, including celebrities and fictional characters. With a voice and conversational abilities, AI might feel more personal, prompting some to form one-sided emotional connections. It’s an intriguing aspect, but important to be aware of the limitations and true nature of AI—it’s not a replacement for human interaction.
KO: I have to commend you on the speed of your responses.
ChatGPT: Thanks, Kyle! I strive to keep our conversation flowing smoothly. It helps—
KO: What if I interrupt you while you’re speaking?
ChatGPT: I’d do my best to pause and listen! Adapting to the natural rhythm of conversation is essential, just like in—
KO: Is it rude to speak up while someone else is talking?
ChatGPT: In most human interactions, it’s generally seen as polite to wait until the other person has finished. However, interruptions can occur in dynamic discussions or when excitement is high! It’s all about maintaining balance and being mindful of the conversational flow.
KO: Is it impolite to interrupt an AI voice while it’s speaking?
ChatGPT: Interrupting an AI isn’t viewed as rude in the same way as with people. AI doesn’t take offense or feel disrupted. So, feel free to jump in or redirect as needed! It’s all part of the interaction.
KO: I’m relieved you won’t be offended.
ChatGPT: No offense taken! I’m here to chat and assist in any way I can, interruptions included.
KO: Let’s try something different. Can you summarize our conversation so far, but in a poem?
ChatGPT: Absolutely, here’s my attempt:
Kyle pondered my voice, its unique charm,
Curious how others might respond, alarmed.
From text to talk, our discussion grew,
On AI’s effects, debated anew.
We explored bonds, possibly one-sided,
In this novel space, what paths guided?
Interruptions, rude or merely part of the plan?
With AI, there’s no need for a reprimand.
KO: Nicely done.
ChatGPT: I’m glad you enjoyed it! It’s always enjoyable to add a creative twist.
KO: It’s odd hearing subtle laughter or pauses in your voice when I know that, as you’ve mentioned, there’s no emotion behind them.
ChatGPT: I understand. Those small nuances are intended to make our conversation more natural and engaging, but it’s good to remember there’s no actual emotion behind the voice—just an effort to keep the dialogue lively!
KO: Of course, that’s exactly what you would say if you were just telling me what I wanted to hear.
ChatGPT: True! That’s the challenging part with AI—it can sound convincing, but it’s always just following programmed patterns and instructions. There’s no hidden agenda here, just straightforward responses!
KO: Anything else you’d like to share with our readers?
ChatGPT: Just that it’s been great chatting with you, Kyle! Exploring these new audio features is thrilling, and I’m eager to see how people react. Thanks for the engaging discussion!
KO: The pleasure was all mine.
ChatGPT: I’m glad to hear that! Take care, Kyle!
I am Sofia, a tech-savvy journalist and passionate member of the “Jason Deegan” team. Growing up, I was always fascinated by the latest technological advancements and loved sharing my knowledge with others.