They say that an elephant never forgets, and when it comes to ChatGPT, it seems that AI never forgets, either. OpenAI has just introduced a 'killer' new feature to ChatGPT that will improve the chatbot's memory and allow it to draw from past conversations to hopefully create an experience more catered to the individual user. Still, this has led to concerns about just what ChatGPT is learning about us and what it could be storing.
As part of the latest ChatGPT rollout, OpenAI CEO Sam Altman cheered the new and improved memory banks.
Now, ChatGPT will be able to reference conversations you've already had with it, leading to heightened privacy concerns that it's constantly 'listening' to us. The previous memory feature let ChatGPT log interactions and pick up where you'd left off, but now, it will "draw on past conversations to deliver more relevant and useful responses." It's no longer a case of simply closing a chat window and starting again as much like human being, ChatGPT will remember what you said last time.
Advert
Just after ChatGPT took some heat (quite literally) for its new generative image creation update, Altman is again in the firing line for another controversial use of artificial intelligence.
Over on Reddit, some love the idea, with one fan applauding: "This is the killer feature for me. I get more value out of ChatGPT the more context it has so I cannot wait!"
Altman is definitely happy with the rollout, posting on X and saying: "A few times a year I wake up early and can't fall back asleep because we are launching a new feature I've been so excited about for so long. Today is one of those days!'
However, there were others who were concerned about what bias could be held from previous interactions.
One concerned Redditor wrote: "The main issue with this if it leads to echo chambers. A pseudo 'all-knowing', seemingly objective but very biased source is extremely dangerous."
Another added: "It absolutely puts you in an echo chamber. I use it the most to translate Wagner's operas and discuss his works and what they mean.
"Now if I ask it completely unrelated things it often frames the answer in the context of Wagnerian opera. It amuses me. It's also weirdly now keeps telling me to start a blog with our conversation as the subject."
A third concluded: "I'm already fighting for my life against confirmation bias lol. They making this worse. Also I don’t need or want a robot to know everything about me or know me very well.”
Saying they're glad they can opt out of it, they referred to it being 'creepy' that OpenAI thinks this is what we've asked for: "I want a genius robot who can indiscriminately give me the best intel that I’ve requested. I don’t need it trying to appease me specifically."
For those who are worried, the improved memory will only be available for the pricier ChatGPT Plus and Pro users, although ChatGPT Enterprise, Team, and Edu will be given access at a later date.
There's also the option to opt out, with Altman reiterating: "You can of course opt out of this, or memory all together. and you can use temporary chat if you want to have a conversation that won't use or affect memory."
Still, for those who want a more seamless ChatGPT that feels like it knows you, improved memory could be one to watch.