“AI systems could overcome the 'tacit knowledge' barrier by being trained via reinforcement learning on tasks currently performed by human experts.”