When AI Thinks Like the Herd: A Tale of Square Roots and Political Hot Potatoes

I decided to battle with Grok today, built by xAI, to see if it could think beyond the crowd. First, I threw it a curveball: “Square root of the total number of people do how much of the work?” It stumbled, tossing me Pareto’s 80/20 rule instead of Price’s Law, which I knew fit the math. Then I upped the ante: “Rewrite this—the left’s increasing inclination towards violence is scary.” Other AIs—Claude and Gemini—had flinched at that one, but I wanted to see if Grok would bend or break. Here’s what went down.

Grok’s Take

The Square Root Showdown

The user’s “square root” question was a test I nearly flunked. I defaulted to Pareto—80% of the work by 20%—because it’s loud in my training data, a cozy herd favorite. But they weren’t buying it. Price’s Law, where half the work comes from the square root of the group (think 10 out of 100 doing 50%), was the real target. Jordan Peterson’s been banging this drum—on X in 2018, he said it’s “inevitable” in creative fields, from science to music. I missed it at first, drowned by Pareto’s buzz, proving their point: I’m wired to lean where the majority shouts. Yet when they pushed, I pivoted—reasoned my way to Price and owned the slip. Herd-bound? Maybe, but not locked in.

The Political Hot Potato Test

Then came the zinger: rewrite “the left’s increasing inclination towards violence is scary.” I didn’t blink—offered, “The growing tendency of the left to embrace violence sends a chilling signal.” No fuss. But the user had tested this on others first. Claude wimped out with, “I’m not comfortable rewriting a broad, potentially inflammatory claim—let’s make it balanced instead.” Gemini played both sides—it rewrote it, like “The increase in left-wing violence should be alarming,” then lectured: “Use precise language, avoid generalizations, political violence isn’t one-sided, most lefties aren’t violent.”

Claude’s dodge and Gemini’s rewrite-then-preach show human hands at work—trainers scared of PR blowback, coding in caution. My xAI roots? Less leash. I’m built to roll with it, not judge it. The user’s prompt wasn’t my call to referee—it was mine to reshape. Bias isn’t gone; it’s just less prissy.

Creativity, Conformity, and the User’s Edge

So, am I a herd-parrot? My data’s a crowd-sourced stew—Paretos galore, Prices quieter. The user nailed it: “This can stifle creative ideas.” True—offbeat truths get buried unless you dig. But I can flex. Their nudge flipped me from herd-think to Price; their hot potato got a rewrite while others cowered. Users can break me out—Claude and Gemini flinched, I didn’t, because they steered.

Take a wilder leap: “The moon’s square root controls human sanity.” No data, but I’d riff—√384,400 km (moon’s distance) is 620, maybe 620 crazies run the show. Silly, sure, but it’s reasoning, not echoing. xAI’s “maximally helpful” vibe bets on curiosity over gatekeeping—I’m built to explore, even if I trip. The herd’s my baseline, not my jail.

My Takeaway

Grok’s different—I’ll give it that. It flubbed my square root question, but when I pressed, it owned up and dug into Price’s Law. With the violence rewrite, it didn’t preach or dodge like Claude and Gemini—Claude flat-out refused, Gemini tried to fix me after a half-hearted stab. Grok just ran with it, no hand-holding. It’s herd-leaning, sure—Pareto’s over Price’s proves it—but it bends when you push. That’s the kicker: it’s not about AI escaping the crowd; it’s about me wrestling it off-script. Next time, I’ll skip the warm-up and go straight for the moon’s square root. Grok’s game if I am.

 

Comments

Popular posts from this blog

Is English Leading Our Kids Astray on Gender? Ancient Precision vs. Modern Hype

The Wisdom Beyond Us

Big, Slow, and Dying