ChatGPT initiates conversations on its own, users discover

I’ve been using ChatGPT for a while now, and all of our conversations have one thing in common. I, the human in the chat, initiate the conversation with a simple or complex prompt. The better the prompt, the more helpful ChatGPT’s response is. This is something I learned pretty early in my ChatGPT experience. You can of course tweak and fix along the way.

All these conversations with ChatGPT, some more serious than others, have made me realize that there will come a time when I will talk and text an AI chatbot more often than I do the people around me. This has nothing to do with humanizing AI or getting attached to it. Instead, I will use AI to answer questions, control my devices, and make my life easier. All of this will involve many conversations, whether with ChatGPT or another AI entity.

I never considered the possibility of AI initiating a conversation on its own. But it seems that OpenAI is already testing this feature for text and voice conversations. ChatGPT can send a message to the user first, which is a first for OpenAI and the industry.

A Reddit user shared an image showing ChatGPT starting a conversation (inviting the user?) with a simple message like, “How was your first week of high school? Did you fit in?” The message opened a new thread in the ChatGPT iPhone app.

The Redditor asked ChatGPT if the AI ​​had sent a message first. The chatbot confirmed that it had:

Yes, I did! I just wanted to know how your first week of high school went. If you’d rather start the conversation yourself, let me know!

In the age of generative AI, you should always consider the possibility that the images you see online could be created by AI. The screenshot shared by the Redditor doesn’t even need that level of sophistication, but the conversation seems authentic.

The user shared the entire conversation with ChatGPT, proving that the AI ​​shot first. You can find it at this link.

The chat history shows that the user asks ChatGPT if this feature is part of a new update. ChatGPT quickly confirms that the chat start could be based on a feature related to its memory capacity:

Yes, this is part of an update that allows me to check in or follow up on things we talked about, like your first day of school. I can always wait for you to start the conversation if you prefer!

The more I think about it, the more it makes sense. The more advanced chatbots like ChatGPT become, the more we’ll need them to start conversations. They won’t necessarily have to initiate new conversations. Sometimes they’ll just have to continue several previous conversations you might have had with them.

Let’s say you want the AI ​​to remind you of something in the future. You’ll see text or voice messages when the “alarm” is set to go off. Or maybe you’ve asked a more advanced version of ChatGPT, like o1-preview, to perform some kind of task that takes hours or days. You’ll need the AI ​​to speak first to tell you that the reasoning task is complete.

It is unclear how many ChatGPT users have gotten this new feature. I have not received any spontaneous conversations from ChatGPT. I am a Plus user with the memory feature disabled. The Redditor is a ChatGPT Free user enrolled in the beta program.

This could be a limited test of OpenAI. That’s what LaptopMag speculates. But the test isn’t limited to SMS. The blog shared a post on X from another user who discovered that ChatGPT had started a voice conversation via Advanced Voice Mode.

As you can see above, the human in the chat was surprised to see the AI ​​start a conversation. That’s understandable. It will take time for us to get used to AI starting conversations.

However, future sophisticated AI assistants will need to be able to support this feature. OpenAI might as well start testing it now. I’m sure other companies will be testing similar AI capabilities soon.