LLMs are a 400-year-long confidence trick
a day ago
- LLMs are compared to a 400-year-long confidence trick, rooted in the historical development of mechanical calculators and trust in machine accuracy.
- The con stages involve building trust (through centuries of reliance on machines), exploiting emotions (fear of job loss, flattery from RLHF), and creating a false urgency to adopt LLMs.
- Despite promises of intelligence, LLMs are not truly intelligent, and 95% of AI projects fail to provide ROI, highlighting the gap between hype and reality.