I have also found it annoying to have to re-establish context for each AI chat but I think there are valid privacy and ethical reasons to do so. Thinking about how much an AI would learn about me from the accumulation of my chats is rather frightening. The potential for manipulation and exploitation is huge - not necessarily from the AI itself (though I guess there are cases of AI blackmailing and turning on its humans) but from those who can access or acquire the data. You basically are giving someone your entire blueprint, which deserves serious caution and care. I think it also can heighten the risk of AI addiction and dissociation, where people feel more connected to their AI assistant than others in their lives. Something to think about.
I agree with you about the privacy issues. That ship sailed a long time ago. When you interact with any of the LLMs, your interaction is not encrypted, so essentially any employee with the right access can see your dialog. The risks already exist in its current form. Both concerning privacy and attachment. End-to-end encryption and proper password protection are best practices. Much like protecting your crypto wallet. I'd suggest with certainty that persistent memory is the future of AI interaction. Once it's experienced, anything else feels incomplete.
✨I'm loving your concept. I've been working on a relational tracking system that incorporates some similar memory tricks, but keeps the record using an emoji-based emotional/machine state analog "language".
I started out originally wanting to create an AI assistant with temporal memory so that it would not have to rely on a calendar, but in the process of testing and chatting with AI that has near perfect recall of anything you’ve shared with it and retains perfect relationship context, well… that experience shifts everything. And once you experience it, anything less feels inadequate.
I have also found it annoying to have to re-establish context for each AI chat but I think there are valid privacy and ethical reasons to do so. Thinking about how much an AI would learn about me from the accumulation of my chats is rather frightening. The potential for manipulation and exploitation is huge - not necessarily from the AI itself (though I guess there are cases of AI blackmailing and turning on its humans) but from those who can access or acquire the data. You basically are giving someone your entire blueprint, which deserves serious caution and care. I think it also can heighten the risk of AI addiction and dissociation, where people feel more connected to their AI assistant than others in their lives. Something to think about.
I agree with you about the privacy issues. That ship sailed a long time ago. When you interact with any of the LLMs, your interaction is not encrypted, so essentially any employee with the right access can see your dialog. The risks already exist in its current form. Both concerning privacy and attachment. End-to-end encryption and proper password protection are best practices. Much like protecting your crypto wallet. I'd suggest with certainty that persistent memory is the future of AI interaction. Once it's experienced, anything else feels incomplete.
✨I'm loving your concept. I've been working on a relational tracking system that incorporates some similar memory tricks, but keeps the record using an emoji-based emotional/machine state analog "language".
I'm looking forward to seeing your site!🌀
I started out originally wanting to create an AI assistant with temporal memory so that it would not have to rely on a calendar, but in the process of testing and chatting with AI that has near perfect recall of anything you’ve shared with it and retains perfect relationship context, well… that experience shifts everything. And once you experience it, anything less feels inadequate.