Now ChatGPT Can Forget!
Yet another step from OpenAI to navigate privacy issues that threaten AI models of the future
When you have powerful opponents questioning your very existence, the need for more allies who swear by your work becomes non-negotiable. And OpenAI appears to be bending over backwards to tell regulators and doomsday predictors that ChatGPT isn’t all that bad and can run on a self-regulatory framework.
Towards this end, the creators have now announced that users would now be able to tell ChatGPT what to remember and what to forget from future conversations – something that one definitely cannot do while conversing with a fellow human. OpenAI started rolling out a set of memory controls today for a small chunk of free and paid users of their GenAI-based chatbot.
What’s changed?
In a blog post, OpenAI says a broader rollout would happen soon but hasn’t actually specified a date for the shift. And what do the new controls do? They allow you as a user to tell ChatGPT explicitly to remember something. You could also check what all it remembers from past chats and then turn off its memory completely.
(Source: OpenAI)
“We’re testing the ability for ChatGPT to remember things you discuss to make future chats more helpful. You’re in control of ChatGPT’s memory,” is how the blog post itself describes OpenAI’s efforts. It says the effort is directed towards testing the memory of ChatGPT to remember things one discusses across chats as it saves one from repeating the information while making future chats more helpful.
However, it takes pains to point out that now the user would be in control of ChatGPT’s memory. “You can explicitly tell it to remember something, ask it what it remembers, and tell it to forget conversationally or through settings. You can also turn it off entirely,” says the post while noting that it was a beta test to learn how useful the feature was.
How is making a difference?
“ChatGPT can now carry what it learns between chats, allowing it to provide more relevant responses … As you chat with ChatGPT, you can ask it to remember something specific or let it pick up details itself. ChatGPT’s memory will get better the more you use it and you’ll start to notice the improvements over time,” the report says.
Of course, one can easily decipher where OpenAI wants to go with this modification. One could ask the chatbot to remember the tone of voice, the preference of topics, formatting preferences for blog posts as well as languages and frameworks for coding. However, what you can ask it to forget includes medical records.
Also, custom chatbots powered by OpenAI models available via the GPT Store come with their own memories that remember things that one has used it for. For example, the Books GPT would remember the titles one read or the genres that you preferred. However, these aren’t anyway shared with ChatGPT – so no need to be concerned!
What OpenAI is now saying openly is that the memory feature of ChatGPT and GPTs can be disabled at any time from the settings menu. Once turned off, both these bots would not create or use memories. Also, the menu provides users the facility to delete specific data or clear all of it from the memories. However, do remember that deleting a chat from chat history doesn’t erase it – for that you need to delete the memory itself.
OpenAI also reiterates that in the future it would be taking steps to steer ChatGPT away from proactively remembering sensitive information related to health and wellness etc. unless a user specifically seeks to retain it. However, one must also remember that OpenAI uses such data to improve its models, barring those using ChatGPT business.