not sure about this article despite ACM, it's a little a johnny-come-late trying to take measure of something without real grasp. Disqualifier includes that the number of parameters in a NN is not equal to connections, but commonly to connections plus a bias scalar for each node. A NN is not a function, it's characterized as functioning under the universal approximation theorem approximating functions. Finally, as a author one could have actually typed in what's being reasoned about into the Chat-GPT under a 'construct your answers scientifically, ask yourself why and how prior to answering and incorporate the why and how in your answers' as a pretense prompt. The transcript in https://chat.openai.com/share/73210b2e-d57f-44b4-a4ad-24e9fe... is quite contrarty to the letter's argument.
