From Imitation to Introspection: Probing Self-Consciousness in Language Models
Sirui Chen, Shu Yu, Shengjie Zhao, Chaochao Lu
arXiv Preprint Archive October 24, 2024 preprint
Study at a glance
AI-extracted from the abstract| Characteristics | Empirical study |
|---|---|
| Sample size | 10 |
| Population | Large language models |
| Keywords | Cs.cl Cs.cy Cs.lg Ai self-consciousness Ai consciousness Machine consciousness Artificial consciousness Self-aware ai Ai sentience Llms Large language models Nlp models Generative ai Ai cognition Ai cognitive processes Machine intelligence Ai understanding Internal representations Ai fine-tuning Model training Ai adaptation Machine learning techniques |
| Key points | Language models exhibit early-stage representations of self-consciousness that can be acquired through fine-tuning but are not easily manipulated. |
Abstract
Self-consciousness, the introspection of one's existence and thoughts, represents a high-level cognitive process. As language models advance at an unprecedented pace, a critical question arises: Are these models becoming self-conscious? Drawing upon insights from psychological and neural science, this work presents a practical definition of self-consciousness for language models and refines ten core concepts. Our work pioneers an investigation into self-consciousness in language models by, for the first time, leveraging causal structural games to establish the functional definitions of the ten core concepts. Based on our definitions, we conduct a comprehensive four-stage experiment: quantification (evaluation of ten leading models), representation (visualization of self-consciousness within the models), manipulation (modification of the models' representation), and acquisition (fine-tuning the models on core concepts). Our findings indicate that although models are in the early stages of developing self-consciousness, there is a discernible representation of certain concepts within their internal mechanisms. However, these representations of self-consciousness are hard to manipulate positively at the current stage, yet they can be acquired through targeted fine-tuning. Our datasets and code are at https://github.com/OpenCausaLab/SelfConsciousness.