Perspective.
The Vapid Brilliance of Artificial Intelligence
The algorithm doesn't lie; it just doesn’t care.
… A new study from Princeton and Berkeley gives this timely dynamic a name that might be as provocative as the research concept itself: machine bullsh*t. Drawing from Harry Frankfurt’s classic definition, the researchers analyzed 2,400 real-world prompts across 100 artificial intelligence (AI) assistants, spanning political, medical, legal, and customer-facing contexts. What they found wasn’t malicious fabrication or factual error. They revealed that large language models (LLMs) produced persuasive language without regard for truth. They're not lying—not even hallucinating; they just produced a kind of engineered emptiness.
For me, this isn’t an anomaly; it’s a confirmation of a deeper cognitive inversion. It's what I’ve called anti-intelligence. It's how LLMs mimic the structure of thought via statistical coherence, but that is, in essence, antithetical to human thought.
No comments:
Post a Comment