Simon Blackwell

A two years of work benchmarking LLMs for empathy (see https://hackernoon.com/search?qu...) showed us how hard it is to get an LLM to behave in a way that users percieve as empathetic over an extended dialog. Empathy requires more than just text from a single dialog, it requires broader memory and context coupled with audio and visual outputs. iCendant Empathy is based on research done during...

While doing research to support benchmarking of LLMs for empathy (see https://hackernoon.com/search?qu...), we discovered a gap in support for emotionally laden text to speech. Although Hume and Speechify have both done extraordinary work in this area, we could not find a solution that adequately supported text-to-speech for meditations and dialog responses that were more than just a few...