ChatGPT – Artificial Intelligence gains human-like memory

ChatGPT is increasingly becoming a reliable assistant, not only remembering what you told it about yourself, your interests, and preferences, but applying these memories to future conversations. It's a seemingly small change that could make creative AI more human-like, and perhaps pave the way to artificial general intelligence, a term used to describe a digital brain capable of functioning in a way more similar to a quintessential human.

OpenAI announced limited testing By sharing it on Tuesday, explaining that it is testing the ability of ChatGPT (both the free and premium versions) to remember what you say to it, across all the conversations you have.

With this update, ChatGPT can remember in a natural way, collecting interesting information along the way, such as whether the interlocutor prefers regular coffee or with sugar, or what the user specifically asks him to remember.

The benefit of having memory in ChatGPT is that subsequent conversations with it do not start from scratch. A new question, for artificial intelligence, may come in an unspoken context. In practice, OpenAI reports, this memory will be applied to future queries. If you mention in ChatGPT that you have a three-year-old who loves giraffes, subsequent conversations about birthday card ideas may include pictures of giraffes.

ChatGPT will not only mimic memories of your preferences and interests, but will use this information to deliver more effective performance to the user.

Perhaps for some, the idea that an AI model can remember a series of conversations and use that information to help you is a bit off-putting. This is probably why OpenAI provides the possibility to disable this function in a simple way by specifying “Temporary Chat”, which can be described as causing a temporary memory loss in ChatGPT.

Meanwhile, just as internet search history works, ChatGPT will offer the ability to delete memories through settings. Alternatively, the user can ask ChatGPT to forget something in the conversation flow.

Currently, the feature is available as a test for some users of the free and shared version of ChatGPT, but OpenAI did not say in its post when it will become available to all users. OpenAI is also adding the ability to collect memories to apps that ChatGPT subscription owners can create, meaning developers will be able to integrate this functionality into their own versions of the model. But they will not be able to access the memories stored inside GPT.

An AI model with long-term memory raises more complex issues than a model with a sketchy, at best, picture of past conversations. One issue is clearly the protection of privacy. Since ChatGPT randomly remembers information it finds interesting or relevant to its users, should we worry about our personal data showing up in another user's conversations? Probably not. OpenAI promises to exclude memories from the data on which ChatGPT is trained.

This development could have major implications in the future. ChatGPT already has the ability to generate human-like responses, but its failures and the difficulty in some cases of even remembering how a conversation started shows that we are still a few billion neurons away.

Memories, especially when they include information that comes to you easily during a conversation, can change this feeling. Our relationships with others are largely determined by the experiences and memories we share with them. We rely on this data to shape our communications and conversations. It's how we build connections with those around us. It's possible that at some point we'll feel like something connects us to an AI model that remembers that we don't like spicy food or that we like anything to do with Rocky Balboa.