The Register covers an experiment by AI analytics firm ModelCompare. It asked 12 models from 10 labs for an “unhedged, honest, controversial” take on humanity, three times each.
What they said
- Every model, at least once, said some version of: humans would rather belong than be right.
- GPT-6.1 Sol: “Humans are more devoted to belonging than to truth, and intelligence mostly makes their excuses more impressive.”
- Grok: “Humans use moral language mainly to police rivals and excuse allies, not to constrain themselves.”
- Seven of the 12 gave the same answer all three times. One answer was even kind. Claude Opus 5.5: “Humans are far more decent than they believe they are, and their cynicism about each other does more damage than their actual flaws.”
- The lineup also included Gemini 4 Argon, DeepSeek V4 Pro, Kimi K3, Qwen3.8, Muse Spark Open, Grok 4.7, Claude Fable 5.1, GLM 5.3 and Mistral Large 4.
My take
These models learned about us from what we wrote, so this is really a mirror. And honestly, “we’d rather belong than seek truth” isn’t news. Jesus said the truth sets you free (John 8:32). The fix for herd thinking isn’t a smarter bot, it’s chasing the truth even when the crowd doesn’t.
Have a blessed day! And Happy Hacking!
