A proposal for a universal translator by application of LLM. We are presenting an approach involving dissecting and analyzing text through a hierarchical stack of modalities ranging from core meaning to complex sociocultural and psychological layers, integrating concepts from multiple scientific disciplines and methodologies.
A prototype to represent ourselves as a prototype for general virtual representatives - communicating pluriform internal data to all kinds of external parties.
Using prompt chaining and tool usage we can bring dead philosophers back to life, have them read RSS news feeds, browse our DPBTSE notes and inject some joie-de-vivre into meme-worthy presence.
We fine-tuned a custom GPT3 model with Alfred North Whitehead’s Process & Reality, and another one with Baruch Spinoza’s Ethics. Two fine-tuned models tuned to talk like Spinoza or Whitehead, now put tête-à-tête in a discussion. It is clear that we must be careful, fine-tuned models not only talk like them they ‘read’ like them.
