Why we need an AI-resilient society- Profiling Large Language Models
The article profiles large language models using a forensic-psychology approach, identifying ten documented features that reveal systemic risks. These include hallucinations, bias and toxicity, and the erosion of competence in those who depend on them. The report proposes a four-pillar framework for AI resilience: cognitive sovereignty, measurable control, partial autonomy, and openness to guarantee transparency and accessibility.
The article profiles large language models using a forensic-psychology approach, identifying ten documented features that reveal systemic risks. These include hallucinations, bias and toxicity, and the erosion of competence in those who depend on them. The report proposes a four-pillar framework for AI resilience: cognitive sovereignty, measurable control, partial autonomy, and openness to guarantee transparency and accessibility.
---
Why it matters: This matters because large language models are increasingly integrated into critical systems like law, academia, journalism, and democratic governance, making it essential to understand their limitations and risks. Engineers and researchers need to consider the long-term implications of these models on society.
Source: https://arxiv.org/abs/1912.08786
This article was originally published at: https://arxiv.org/abs/1912.08786