Now ChatGPT can remember (or forget) what you say.
As part of the test, OpenAI deployed advanced settings for managing the chatbot’s “memory” – for a small part of both paid and free users.
“ChatGPT can now carry what it learns between chats, making responses more relevant. During the conversation, you can ask him to remember something specific or allow him to record further details. ChatGPT’s memory will improve the more you use it, and you’ll start to see improvements over time,” OpenAI explains in a blog post.
The update will really make the responses more personalized — for example, if you ask to remember that you have a pet, then probably in travel inquiries or similar, ChatGPT will take that into account.
On the business side, a chatbot’s memory will come in handy, for example, for remembering tone, voice, and formatting options for blog posts, as well as programming languages and frameworks.
Special chatbots created on the basis of OpenAI models and available in the GPT Store have already received their own memories. For example, Books GPT can automatically remember which books you’ve read and which genres you like best. But these memories are never transferred to ChatGPT and vice versa.
The memory function for ChatGPT and GPT can be turned off in the settings menu. It is also possible to erase individual memories or clear the memory completely (but, note, it is about the memory, not the history of requests).
We’re testing ChatGPT’s ability to craft things you discuss to make future chats more helpful.
This feature is associated with a small portion with free and short users, and it is easy to release or off. pic.twitter.com/BsFinBSTbs
— OpenAI (@OpenAI) February 13, 2024
OpenAI says it is taking steps to protect ChatGPT from “proactively” remembering sensitive information, such as health information, unless the user requests it.
In addition, the company will launch a temporary chat option (also initially in the test), which will allow you to start a dialogue from a “clean slate” — ChatGPT will not know about previous conversations and will not gain access to memory, but will carry out special instructions if they are enabled. At the same time, OpenAI says it can keep a copy of temporary chat conversations for up to 30 days for security reasons.
Meanwhile, OpenAI CEO Sam Altman compared ChatGPT to a “barely useful cell phone” during a virtual speech at the World Government Summit.
“We don’t see as many world-changing programs as we would like,” Altman added.
The head of OpenAI says that in terms of comparison with the evolution of mobile phones, he would like to see ChatGPT in the form of the latest iPhone model, but in the world of AI.
According to Altman, OpenAI is currently working on the next big AI language model, GPT-5, which is expected to be “smarter than ever.”
GPT-4 was released back in March of last year, followed by an improved version of GPT-4 Turbo — the most advanced AI model, the launch of which OpenAI publicly acknowledged. There were also rumors about a “secret” release of GPT-4.5.
Meanwhile, OpenAI’s CEO is asking the world to “be patient” and says that GPT technology will get “a lot better in a few years.”
Join the competition of ITS authors! Win cool prizes from our Logitech partners – professional gaming steering wheel and low-profile gaming keyboards.
