ChatGPT surprises users with personal questions: Bug or new functionality?

ChatGPT, the artificial intelligence chatbot developed by OpenAI, started to initiate conversations with its users completely unexpectedly.

context

A Reddit user posted a screenshot that quickly went viral. In this image, ChatGPT appears to start a conversation on its own by asking the following questions:

“How was your first week of high school? Have you settled in well?”.

Everything happened without any prior request. When the user asked if the chatbot had actually messaged him first, the AI ​​confirmed, explaining that it just wanted to check how things went.

Online debate

This incident generated a heated debate online. Some have speculated that OpenAI could be testing new functionality that allows the assistant to initiate conversations, possibly as a strategy to increase engagement.

Others have linked this behavior to the recently launched new ‘o1-preview’ and ‘o1-mini’ models, which are said to have a ‘human-like’ ability to ‘reason’ and tackle ‘complex tasks’.

Patch issued by OpenAI

The company has acknowledged the existence of a bug and announced that it has issued a fix.

According to their explanation, this situation occurred when the model tried to respond to a message that had not been sent correctly and it appeared empty. As a result, it either provided a generic response or relied on ChatGPT’s memory.

Reports of similar behavior

Other Reddit members reported similar behavior: “Same thing happened to me this week!!” or “I asked him some questions last week about some health symptoms I’ve been having. And he texted me this week asking how I’m doing and how the symptoms are going!! It scared me terribly”.

Possible explanations and controversies

While some publications supported the authenticity of the exchange by examining the log on ChatGPT.com, AI developer Benjamin de Kraker demonstrated in a video that a similar effect can be achieved by manipulating custom instructions and manually deleting the first message.

It seems that OpenAI is trying to make chatbot conversations something natural and personal, similar to their efforts for Voice mode.

Concerns

While the idea of ​​a ChatGPT initiating conversations may seem harmless or even useful, it raises privacy and autonomy concerns.

Source: MSPowerUser