“Gavin Baker posits that the true intelligence of each new generation of AI models may never be fully known because new, more capable models are released before the previous ones can be thoroughly evaluated.”