[bugintheprogram]

I'm not really speaking to the utility of the word, which I think you're more getting at, but more of the hypocrisy of the stated reasoning. He uses clanker to "prevents anthropomorphizing" AI but by having an insulting name, it implies humanizing it (because insults are used to hurt the insultee's feelings or dehumanize them, and to do so implies first humanizing them to have feelings to be hurt by said name). Calling an agent a "conformal prediction large language model" or similar better fits his argument because by it being a description of what it is, it doesn't inherently humanize it in the same way an insulting name must by its nature.