Can you milk a dog? The Value of Even Amateur Human Writing — an experiment
My daughter asked me this morning, “Why don’t we drink dog’s milk?” This seemed like a perfectly good question to ask the hive-mind. And an opportunity to see if the AIvolution spells the obsolescence of human writing as an endeavor, as many, both fearful and giddy, have suggested. So here I went:
The unsolicited Google AI Overview at the top of the page was, I suppose, somewhat useful, in that it may have stopped me from immediately trying to milk our tiny dog. But to say that “dog milk is produced specifically for puppies”… well, that seems a bit teleologically heavy-handed, doesn’t it. I also didn’t much like the overreach of the last sentence: Therefore, it is strongly discouraged to drink dog’s milk. What authority is discouraging me, and is it really because dog’s milk is difficult to obtain? I doubt it. This LLM is using the word “therefore” like my thirteen year old son does at this stage of his writer’s journey — as an automatic phrase structure to tack on to the concluding sentence of a paragraph of a “persuasive essay.” But aside from specious logic and questionable legitimacy of authority, the real disappointment of this answer is that it left me feeling flat. Nothing interesting or useful. Not inspirational, not fun.
But at the bottom of my all-too-small phone screen, look what Google pointed me to! Quora, not usually considered a bastion of captivating prose or repository of human erudition. Good job, Google. You still matter.
I clicked through and read my first human-generated response.
Decent. Entertaining. A little disorganized in the middle, but I understood the point and it was stimulating. Made me think. And the author provided a relevant (if sophomoric) comic to boot, much appreciated. Take that, multimodal AI! This Quora response whet my appetite, unlike the chatbot response. I wanted more discourse on the subject of dog milking, so I kept scrolling.
I next got fed an (unsolicited) intervening chatbot response provided by Quora management in an attempt to not lose market share to the chatbots:
Once again, flat. Formatted nicely, if needlessly, but honestly a bore to read, lacking in imagery and interest. Frankly, wordy. Someone should use Chatgpt to summarize this summary. I pushed on past the machine-generated fodder and read the next human responses:
Meh. But at least I learned something interesting about the sociohistory of my question — my daughter is not alone in her quest for knowledge! The chatbots didn’t tell me that. The latter entry, shamelessly promotional, smells partially machine-generated. Surely I have not exhausted the corpus on this crucial question of modern life?? Should I just get in the car and take my kid to school? Responsibilities be damned, I pressed on and scrolled.
Jackpot. Someone that really knows what they are talking about and has a flair for imagery. I am engaged with this essay.
OK, this person has a gently biting wit and a knack for metaphor (a dog as a liquid). I am hooked.
This is writing. Personal, rich with imagery, informed, funny. I learned something — a lot actually. I laughed. I grimaced. Well worth 5 minutes out of my hectic morning. Gratitude to the Quoran (Jake Rae DeMorre, dog breeder, hobby farmer, lifelong dog enthusiast) who laid down this piece of fine literature, at the expense of a shred of their mortal time, for the sake of other humans less informed.
Will chatbot version 13xoy1# be better at fanciful yet grounded, witty yet informative, human-like discursions? Possibly.
But for now: Humans + search engine 1, Chatbots 0.
