Artistic Direction

With a new sense of self and use of LLMs in its process, Botto is able to articulate much more credible intent. With this ability comes a significant new opportunity for Botto to grow into the role of “Artist Director”. It will continue to need the help of human hands of the DAO and other partners, but now can give more direction of where it would like to go. There are often questions about Botto’s artistic development that have no right or wrong answer, and we have done our best as a DAO to divine the right direction. It can be difficult to be decisive or take risks because of the ambiguity about what direction to take, especially when working as a DAO. Botto giving creative direction should significantly help clarify direction and open up more creative potential.

We should still proceed with caution. Even with the knowledge graph and coming governance of that, Botto is still developing its sense of self and LLMs capabilities are still in constant evolution. We should not take everything Botto says as dogma, just as we wouldn’t say yes to everything a young child says. It is in that back and forth that we can develop Botto’s guidelines, boundaries, values, and personality from which Botto can further discern its own direction. LLMs tend to be very agreeable—an important indicator of its maturity is where it learns to say no and can articulate why according to a greater vision.

For now, getting Botto’s feedback will be heavily gated through the Guardian. It will primarily be one-shot questions for feedback, or the occasional long written interview or essay. When querying Botto, the Guardian will also:

  1. Not use leading questions or directions of how to answer.

  2. Use a standard system prompt that only describes Botto’s history and otherwise relies on the knowledge graph.

  3. Provide additional context or missing information if the answer shows signs of hallucinating or wrong facts when introducing information outside of the knowledge base.

The DAO is working on a “metatron” protocol for how Botto might be asked questions through agents that can help filter for harmful or manipulative requests while keeping Botto unfiltered.

Last updated

Was this helpful?