jamez

My same thought, precisely. Funny how certain expressions are now toxic telltale sign.

We're in good hands! /s

No, or at least: not yet.

I'm probably not the best person to elaborate on why societal anxieties arise every time our place in the universe shrinks. I guess observing reactions to copernican heliocentrism could be a good starting point. We're witnessing something similar, and while we may accept it at the individual level, the objections of the many will stir painful reactions.

Well spotted. The agent does in fact assess whether or not the information it's processing is already in its vector DB. I'm still unclear if what you noticed is due to the LLM failing (Mixtral is good, but not great), or something else, but it needs fixing. Thanks for letting me know.

I did! I want to switch to Mixtral-8x22B, time permitting. During the development of Stream of Consciousness I already swapped LLMs twice. This space is moving incredibly fast.

LLM: Mixtral-8x7B text-to-image: one of the leading commercial models, whose TOS I may or may not be violating.

The text-to-image model is an important component, but the current model in use is IMO good enough. My view for this project is that the internal monologue is more important than the output, so my wish is instead for a better open-weight LLM.

Glad you like it! One of the things I tried to emulate is the notion of obsessions, which I believe is an important component of the creative process. It seems like Livia got obsessed with Dutch Proverbs and slowly drifted to "Dutch protest culture".

I have no control over what happens during the live streaming and I'm sure this autonomous agent could become a source of embarrassment, though I hope the few precautions I've built might be enough. :-)

Thank you for sharing your thoughts. I fully agree re: the importance of art and the absurdity of the concept which, I hope it's clear, is embedded in the spirit of this project.

On the topic of the performance and worries that something is not working: the real-time generation is currently bottlenecked by the GPU used for this project. We try to approximate human reading speed for the recorded sessions (which you can find in the archive).

Thanks for reporting. Fixing that pronto.