I do not have time right now to view the video, but here is a tangential response: LLMs and machine translation have reached the point where they can do all of the tasks required to implement a Chinese Room - not perfectly or even undetectably, but to the point where at least I would regard it as tendentious to insist the gap can not be closed. Given that, it seems difficult to insist that the the human-operated Chinese Room is impossible in principle.

Of course, Searle did not say that implementing a Chinese Room was impossible, but if it is possible, then it seems he must fall back on the position that it does not need understanding to work. That might be the case - whether current AI understands anything is debated today - but if it does not, then Searle simply chose the wrong sort of task to make his point; he might as well have chosen an electronic calculator / optical scanner combination and asked whether it understood arithmetic. This shows, I think, how much the argument, when it was presented, depended on pumping anti-computationalist intuitions shaped by the technology of the time, and it needs a different, more difficult task to remain relevant today.