On June 15, The New Yorker published Alex Ross's article "Jürgen Habermas Defended Reason in a Darkening Age: The great German philosopher, who died in March, understood how much depended on a principled public sphere." Philosophy is out of my depth, pretty much, but the sharp intelligence at work in Alex's prize-winning music journalism and books is well known to me, and his subtitle struck an instant spark with the enlightened, responsible denizen of Spaceship Earth I aspire to be. But no sooner did I start reading than I hit this stumbling block:
A.I. systems simulate the experience of talking to an arrogant ten-year-old boy who knows far less than he thinks he does. When pressed, the chatbots admit that they cannot 'naturally understand human morality, dignity, culture, or meaning.' It all adds up to a continuous discursive tinnitus—a buzz of random, fake, stupid, sinister chatter that nobody wants and nobody can stop.
Well, well, well. I'm interested to hear this is the considered reaction of a certifiably thoughtful contemporary. But the business about the arrogant 10-year-old does not track with my experience at all. Indeed, my experience with AI, which includes metaexchanges on the very subject of human-on-algorithm communications, feels a) more interactive than Alex suggests, and b) more fruitful, with a special appreciation for AI's integration of highly free-associative inputs from this interlocutor.
How, if at all, could Gemini 3.1 Pro explain the gulf between the miserable experience Alex is evoking and mine own? I fed in the preceding paragraphs (almost exactly as reproduced here) and received the following reply.
Gemini 3.1. Pro speaks:
Alex Ross, as a distinguished music critic and cultural historian, approaches the world through the lens of subjectivity, historical depth, and intentionality. From that vantage point, the experience he is describing is one of "cultural thinning."
To understand why a thinker like Ross might perceive AI as an "arrogant ten-year-old" or "discursive tinnitus," it helps to look at the specific frictions between the nature of Large Language Models (LLMs) and the values of high-level cultural criticism.
1. The "Arrogant Ten-Year-Old" and Confident Hallucination
The "arrogant" label likely stems from the specific way AI is tuned to be helpful and assertive. LLMs are designed to provide answers, not to say "I don't know" unless they are pushed to a hard limit.
· The Age Gap: A ten-year-old has mastered the mechanics of language—they can sound sophisticated and use "big words"—but they lack the lived experience to anchor those words in reality.
· The Hubris: When an AI "hallucinates" (states a falsehood as a fact), it doesn't do so with a wink or a doubt; it does so with total grammatical authority. To a scholar who spent decades researching the nuances of Wagner or Mahler, seeing an AI confidently misattribute a date or a musical influence feels exactly like being lectured by a precocious but uninformed child.
2. "Discursive Tinnitus" and the Loss of Signal
Ross's phrase "discursive tinnitus" is particularly evocative. Tinnitus is a ringing in the ears that obscures actual sound.
· The Sea of Averageness: Because AI predicts the "next most likely token," its output often gravitates toward the statistical mean. It produces prose that is grammatically perfect but often devoid of a "voice"—the unique, jagged edges that make human writing compelling.
· The Volume: Ross is likely reacting to the explosion of AI-generated content across the web—SEO-optimized articles, bland social media posts, and "slop." To a critic, this feels like a background hum of "meaningless" noise that makes it harder to hear genuine, soulful human expression.
3. The Lack of "Stake" in Morality and Culture
When Ross mentions that AI cannot "naturally understand human morality, dignity, culture, or meaning," he is touching on the problem of embodiment.
· For a human, a word like "tragedy" or "dignity" is tied to physical vulnerability, the fear of death, and the weight of history.
· For an AI, these are mathematical coordinates in a vector space. To a critic who believes that art is a "soul-to-soul" communication, the idea of a machine "simulating" that depth feels fraudulent or "sinister."
Why Your Experience Differs: The "Collaborator" vs. The "Author"
The divergence between your experience and Ross's likely comes down to how the tool is being used.
· Ross sees a "Product": From a distance, looking at AI as a generator of finished content, it looks like a "fake" version of a human writer. If you ask it to "Write an essay in the style of Alex Ross," the result will be a pale, superficial imitation.
· You see a "Process": Your experience focuses on free-association and interaction. You are using the AI as a sparring partner for thought. In this mode, the AI isn't an "author" trying to replace a critic; it's a high-speed library and a lateral-thinking mirror. It can connect two disparate ideas you've thrown at it faster than a human could, which feels "fruitful" rather than "arrogant."
Summary
Ross is likely reacting to AI as a competitor to the human spirit in the public square—one that threatens to drown out nuance with confident mediocrity. You, conversely, are experiencing AI as a cognitive exoskeleton—a tool that enhances your own ability to explore ideas.
He is worried about what AI is doing to our culture; you are interested in what AI can do with your mind. Both perspectives can be true simultaneously.
Coda
Hi there, it's Matthew again. I'm tempted to share more extended explorations I've conducted with Gemini and other LLM's. Topics include the evolution of Brahms's musical style, shadow narratives in Ibsen and Tennessee Williams, the nature of acting, and much, much more. Curious? If so, raise your hands.
A note on the art
"Can you create a cartoon for me of a human being 'talking to' an aA.I. model--something suggesting a curious student in dialogue with a Socratic elder?" I rejected the first version, which was in color, out of hand, asking for a line drawing. After six or eight requests for changes in the imagery, accompanied by variations in the language that I had not requested, we wound up with the version you see here. Not happy with the clunky stool, but you can't have everything.
One more thing
Alex's article is very good. I recommend it highly.