AI Hits - ChatGPT Combo of Memory and Live Voice Mode
This morning started with a lovely walk with my dog, enjoying a warm, sunny Spring day and the big blue Texas sky. Back indoors I was thinking on what to write about today. When we got home I had a fun little chat with ChatGPT - which ended up being a perfect reminder that I have been meaning to share thoughts on this combo of features.
By now there are lots of great individual features in the GenAI apps that I use regularly. Claude and ChatGPT are often the best across many different tasks, Gemini is a little hit and miss, but getting better, Perplexity is excellent at web search, Mistral’s Le Chat is another good all-rounder.
But ChatGPT’s combination of ‘memory’ and live voice (officially called advanced voice mode) is unique in my experience. The memory feature does what it sounds like - it remembers things that you ask it to, and also at least some past prompts that seem relevant to the core set of my interests I’ve shared with it. Live / Advance Voice mode gives us the ability to have a continual conversation with ChatGPT.
My quick conversation with ChatGPT this morning was a fun little test that shows off this combo of features, with its Live Video feature thrown in as well. Here’s the beginning of our chat:
The cool thing here is that ChatGPT mentions my dog’s name, gets it right. While we continued the chat I opened the camera app on my phone and pointed it at a bookshelf, and asked ChatGPT could name a couple books that are related to my job - so it needs to remember that I told it what that is months ago. Here’s a look at that part of our convo:
The first book ChatGPT named, Tribe of Hackers, is correct, the second one was a hallucination, not a book that exists on my shelf. It got it right when I told it that though and asked it to name another. Despite the stutter, the main point is that it knows I work in cybersecurity and is looking for books on that in the shelves.
This is a combo of features that is fun to use, but they’re more than just fun. They are also time savers. A continuous voice conversation can be a great way to move beyond single prompts or trying to create a perfect prompt, because we can course correct ChatGPT and flesh out more depth and detail as the conversation goes on. It’s usually much faster to use my voice to enhance the interaction, rather than typing.
Memory offers similar time saving advantages. I now have a number of areas that interest me, some work related, more that are life related, that ChatGPT is aware of, remembers. So I can say “Hey, remember last time we talked about healthy eating” and its recall is spot-on and it will tell me (correctly) that last time we talked about that it was looking at combining Mediterranean and plant-based eating habits. I can start with the same sort of voice or text prompt and ask about our last discussion on PKM and notes apps, or Buddhism, and so on.
All of this is in the realm of small time savings that become more substantial the more we use them.
In all of our use of GenAI tools, we should be careful and mindful of what we share in our interactions with them. This is certainly true with the memory feature in ChatGPT. ChatGPT has settings that support the cautious use of this feature. For starters, it is Off by default - you need to turn it on under Settings > Personalization:
You can also delete all of its memory by clicking on Manage memories and then “Clear ChatGPT’s memory”.
Are any of you using ChatGPT’s memory and live voice features? If so, what do you think of them, how are you using them?
I am and remain impressed with voice and think with another turn of the crank it'll be increasingly how we interact with AI. Both sides of the can and string - AI understanding our speech, and responding back quickly and coherently are amazing.