OpenAI is currently testing ChatGPT with memory. Editor Mike believes that while this news is useful, it also raises concerns about AI starting to resemble dystopian science fiction scenes. It’s worth considering whether this is something we should be worried about.
ChatGPT gains memory: what does that mean?
It’s been inconvenient that ChatGPT didn’t have memory. For example, if you asked for fun things to do in Amsterdam, you’d receive the same answers each time, without the app remembering that you already went on a boat tour. A memory feature would also be practical for personalization, such as when looking for a gift for your daughter. OpenAI is now testing a new version of ChatGPT that is capable of remembering things. This deepens the functionality of the service, shifting it from an anonymous helper to a sort of personal assistant. The memory feature can be expanded in two ways: either by manually entering information or allowing the AI to filter it from your conversations.
According to OpenAI, the memory feature can be entirely disabled, and users can also erase select parts from the memory if needed. Additionally, users have the option to engage in ‘temporal chat’ with ChatGPT, similar to incognito mode in browsers, where the information shared will not be stored in the user’s ‘profile’.
Privacy is a rational concern
The small group of users currently testing ChatGPT with memory have the feature enabled by default, indicating OpenAI’s purpose in introducing these new capabilities. The information shared with the app is used to further train ChatGPT and contribute to the development of artificial intelligence. While users can erase certain elements from ChatGPT’s memory, there is no control over what the company does with this data, requiring a level of trust when it comes to OpenAI and user data.
At present, OpenAI does not sell personal data and preferences to advertisers. However, it’s uncertain if this will remain the same in the future. As AI has financial potential, it’s possible that (the free version of) ChatGPT will incorporate some form of advertising, making it more susceptible to personalized ads.
Science fiction
Rational concerns aside, AI’s ability to mimic humans can evoke an eerie gut feeling. The better AI becomes at impersonating humans, the more discomfort it brings. This leads to ‘The Uncanny Valley’, where it becomes difficult to distinguish man from machine, resulting in discomfort.
As ChatGPT gains memory, it’s becoming more of a ‘real’ presence in our lives. Like friends and family, the app can recollect details shared weeks or months ago, resembling a dystopian science fiction scenario. It’s unsettling when AI knows something about you that you don’t remember sharing and starts giving unsolicited advice.