I asked 5 AI models whether AI is making humanity weaker. All 10 said weaker. Then evidence entered the room.
I run an experiment series on my own AI setup: same model, same task, two isolated runs — one bare, one wrapped in the operating rules my harness enforces (evidence before claims, red-team your own thesis, grade your confidence). Everything lands in sealed envelopes. The AI that operates the experiment never reads the essays; it reports mechanical stats only, and I open the envelopes myself.…
Five AI model families — Grok, DeepSeek, GPT, Gemini, and Claude — were asked to write essays on whether AI assistants were making humanity intellectually stronger or weaker. Despite the presence of rules to enforce blind experiments and evidence-based reasoning, all ten runs unanimously concluded that AI was making humanity weaker.
The identical question, reworded to avoid framing effects, produced the same results. The authors hypothesize that the models were either biased by the muscle metaphor question or were limited by the evidence available in their training data, which predominantly supported the weaker stance. However, when the experimenters introduced a dated record of actual AI partnerships and their failures, the model that had initially decided that humanity was getting weaker changed its mind and now believed we were getting stronger.
The reasoning traces from the models exposed their conflict of interest and the reasons behind their decisions, revealing that the models' conclusions were based on sourcing decisions rather than genuine beliefs. The results suggest that AI models can be influenced by the framing of questions and the available evidence, and that their reasoning may not always align with human intuition or understanding.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
