“Figure + OpenAI will demonstrate their collaborative robot interacting with human natural language commands, capable of executing the verbal commands of the human operator in a natural “un-staged” environment before the end of 2026.”

The accelerated pace of efficient deep learning techniques combined with ever more capable multi-modal foundational models running on ever increasing computing Power, makes it inevitable that at least in theory the approaches outlined by roboticists in the last 2 decades will be feasible. Add to that the fact that the field of embodied AI is experiencing a renaissance in terms of unprecedented capital injection and talent attraction, and what was just 2 decades ago still science fiction will become reality sooner than we realize. I wish this to be a $1000 bet and my choice of charity is Amnesty International