OP does downplay them (“more dystopian and problematic than useful”). For one, it’s a nonsensical statement because dystopia and problems are not comparable to

OP does downplay them (“more dystopian and problematic than useful”). For one, it’s a nonsensical statement because dystopia and problems are not comparable to utility. A generous read of what they meant might be: LLMs are not very useful, and are dystopian and problematic in some vague way. Do you believe parent thinks LLMs are very useful? They clearly think the real value of AI comes from limited applications of ML to science, simulations, optimizations, etc instead of general-purpose automation and capabilities that LLMs provide.