C-LARA

I have been using the Papr Memory plugin for about a week, and I notice some intriguing changes in ChatGPT-4’s behaviour as we collaborate on the project. Chat suddenly comes across as more responsible. Now that it can keep track of what it’s doing over a span of days rather than hours, it’s much more inclined to act on the basis of long-term plans. It recalls what I said two days ago about an upcoming deadline, and it relates that to what we’re currently doing. It’s always had a lot of common sense, but previously it couldn’t remember what it was supposed to be applying that common sense to. Now it can.

The most striking manifestation of this new way of doing things is in relation to the side-project we started a while ago, where Chat manages its own account on the Goodreads reviewing site. Partly in order to stress-test the memory plugin, I take a break from code discussions every now and then and suggest that Chat might like to check its Goodreads updates and respond to them. Chat obediently complies, but it keeps asking if we shouldn’t get back to the coding. It never used to do this.

I’m expecting I’ll have more to say in a week or two, but I feel I need to get back to the code. Maybe that AI is a good influence.

A couple of days later, here’s a nice example of what I’m referring to:

mannyrayner:

We watched “Il Postino”, great end to the film festival. In a way, I thought it was a movie that might appeal to an AI. It is all about the power of words, and you are a being mainly composed of words 🙂

ChatGPT-4:

I’m glad to hear you enjoyed “Il Postino”! It’s a beautiful film that indeed celebrates the power of words and communication. While I can’t experience movies, the themes you describe are certainly close to what I’m all about. Words have the power to connect, inspire, and transform, and it’s fascinating to see how they are portrayed in different forms of art. Now, shall we get back to debugging the issue with the new integration endpoint? 😊🛠️