The current “AI for Good” movement (i.e., the push to fund or deploy LLMs as social impact solutions) should be treated as a danger to society. These same tools power automated weapons and surveillance tech, formed via an environment-destroying training process over non-consensually collected data, and most notably, lack any quality evidence that they are even wanted, let alone helpful.

The Kaleidoscope Project was established to provide a counterbalance. Our work is grounded in the belief that “AI for Good” requires high levels of community authority, robust measurement of both benefits and harms, and a reorientation towards AI tools that aren’t LLMs or otherwise generative.