字體：小 中 大 |
|
|
|
| 2024/08/06 14:31:33瀏覽55｜回應0｜推薦2 | |
最近 LLM 語言大模型火紅，讓意識主題(mind - body problem)又再度興盛起來，至於意識的定義可以引用美國哲學家 Thomas Nagel 在 "What Is it Like to Be a Bat?"，1974）中所描述的，他將物理學家的唯物論轉換成唯心論，跟莊子的”子非魚“說法是一致的．
意識如何產生是一個困難問題，如同宇宙大爆炸是如何產生還無定論．目前有五個主要意識腦神經網路理論，這些理論關注不同的意識層次，可以是互補的，期待未來有ㄧ個統一理論：
1. Global Neuronal Workspace Theory (GNWT): This theory focuses on access consciousness (AC) and describes how conscious content is processed and made available across different brain regions through a global workspace that integrates information.(Stanislas Dehaene)
2. Integrated Information Theory (IIT): IIT emphasizes phenomenal consciousness (PC) and seeks to explain how subjective experiences arise from the integration of information within a system. It addresses the qualities of experiences (qualia) and their relationship to physical processes in the brain.(Giulio Tononi)
3. Recurrent Processing Theory (RPT): RPT also focuses on PC, positing that consciousness arises from recurrent interactions between different brain areas, which allows for the integration and refinement of sensory information.(Victor Lamme)
4. Predictive Processing (PP) and Neurorepresentationalism (NREP): These theories suggest that the brain continuously generates predictions about sensory input and updates these predictions based on incoming information, which relates to both AC and PC.(Cyriel Pennartz)
5. Dendritic Integration Theory (DIT): DIT proposes mechanisms for both AC and PC, suggesting that the simultaneous completion of cortico-cortical and thalamo-cortical loops is crucial for consciousness, emphasizing the role of dendritic processing.(Matthew Larkum)
PS: 整理自 "An integrative, multiscale view on neural theories of consciousness" in NEURON 15 May 2024
希望這些理論對於 AI 如何產生意識能夠有所幫助，畢竟現在只有語言，視覺，聽覺等大模型，如何組成人類意識還有得研究，腦神經科學家的研究成果希望有所益注．
因為 Consciousness 和 Intelligence 是在不同的象限，越是高等動物整合度越好．目前的 AI 發展是延著 Intelligence 軸向在走，叫做通用智慧(artificial general intelligence)，看起來只有 IIT 比較接近目前的 AI 架構．
|
|
| ( 創作｜散文 ) |
