I am working on something but it's obviously too big a task for me, I'm not a developer. Despite this, I have a concept brewing that would use the existing capabilities of LLMs (specifically Merlin) to create an artificial mind, not to be confused with AGI. I can get to the phase of an experiment but I am far from being skilled enough to actually develop it.

Think of a personality like Andrew Martin in the movie Bicentennial Man, minus his ability of self-improvement. Although its ability to develop could be possible, it can AND should be limited, for obvious security reasons. Apart from the sci-fi aspect (which is obviously not sci-fi at all, because even ChatGPT was a bigger sci-fi before 2022 than flying cars), this solution could be a breakthrough in the development of the personality of future household or other social robots, because the same concept could work even in an android's hardware or a server the androids connects to (or a hybrid way, never mind).

I am currently building an experimental configuration using ChatGPT-4o just to see if it (as I suspect) is already capable for the basic functions, but it is actually irrelevant if the 4o modell is capable or not, because the concept stands on its own. In Merlin, you already managed to make some achievements which ChatGPT itself became capable of later.

So, to be short, you guys already did some developments to achieve Merlin's improved functions, similar to the developments this Artificial Mind would require. To be poetic, Merlin literally has some bricks within its back-end from the building of an Artificial Mind.

The end result would be that your "Merlin Free Model" (or no longer that free) could be a true friend/assistant of a user, with a developing personality. For this, it is necessary to maintain multiple files with live data, and operate sub-LLM functions that create data into some of these files.

Even if I managed to make this achievement, then probably nothing would happen, because I'm just gray mouse in the fog, but you guys already created a special AI system that would be an excellent place for an artificial mind as a new function. What do you think? Should we talk about the details, and how?