Here’s a bit of an admission, though hardly a shocking one. Recently I have been selling my soul, or at least a big chunk of my privacy, to GenAI chatbots. Well, giving it away really. No, it’s worse than that - I am (in one case) paying the AI provider for the privilege of serving up my privacy and my soul on a platter.
As soon as I read a great post called ‘ChatGPT’s "Memory" Is Kind of Meh. But You Can Change That’ by Daniel Nest a couple weeks ago I was more than a little intrigued with the idea of trying out this new feature where ChatGPT 4o could “remember” things about me. Daniel’s post also offered some good guidance on how to set Personalization instructions for ChatGPT 4o to help get the best results in its remembering / getting to know me efforts. I put those to use along with some of my previously established custom instructions.
I initially tested out the memory capabilities of several of the GenAI chatbots I use, including Gemini and Claude’s latest versions - but the two I have ended up spending the most time on this with are ChatGPT 4o and Pi.
I don’t recall seeing Inflection - the makers of Pi - talking up any ability for Pi to remember things “said to it” within prompts, but I had got the impression that it could just through my general daily usage of it. It turns out that impression is not wrong. Here’s a quick recent example of that, where I first asked Pi to say whether it recalled a specific cyber threat intel post I had shared with it, followed straight after by a purely personal question about my dog’s name:
ChatGPT 4o offers a lot more in this area. For starters, you can include voice, text, photos and documents to it as part of helping it get to know you. In this instance I used my phone’s camera and some text to show it (even more evidence) that I work in cybersecurity:
And here’s ChatGPT 4.0 remembering more about my dog and being quite a good conversationalist while engaging with it via voice:
I like how ChatGPT makes an entry stating that its memory has been updated in all these sort of interactions with it. And I’m very impressed with the conversational ability that ChatGPT 4o is showing. This is a screencap from just after I had shared a few thoughts on my current favorite book about generative AI, ‘Co-Intelligence: Living and Working with AI’, by Ethan Mollick and sharing that I am an AI optimist:
As you can see, ChatGPT and I are on the same page, looking forward to more conversations about working with AI in co-pilot or cyborg mode (Ethan Mollick’s term). Honestly, I’m looking forward to lots more conversations with ChatGPT and Pi on lots of subjects. Of course I take care on which subjects I engage in and what level of information I provide to these bots. Just about all of my discussions are either around researching topics with prompts that could just as easily gone into a search engine or low level personal stuff about my dog. For me, my protection ship for this level of privacy sailed long ago :)
You got me with the "Antifragile" book shown on the picture of your books. Haven't read the other ones you have ther, but would like to. Which cybersecurity book do you recommend?
Now Ai data grab includes visual info about your house. Each new “development” involves deeper penetration into our lives reconceptualized as data. How do we strike an appropriate balance? Asking for a friend. 😅