Concept Safety: Producing similar AI-human concept spaces — AI Alignment Forum