This morning I had a brief episode of what I have taken to calling "Being out of phase with yourself." It felt as if I was parroting my own actions as I was performing them, as if I was suddenly and briefly cognizant of the delay between my unconscious movements and my conscious awareness of them.
This particular incident led me once again ponder the age old question of Agency versus Determinism: Am I actually in control of my own actions? This led me down a rather fascinating rabbit hole reading work on derealization from Metzinger to Hume to Libet, and ultimately arriving at the traditional contradictory conclusion, Yes & No.
But one new thought did enter my mind that seemed like one worth following. Consider the ideomotor effect and the experiments of Gazzaniga and Sperry on divided hemispheres. The unconscious mind can and does control the body with zero input from the conscious mind, and not just in autonomic nervous system reactions; the hands move on the Ouija board and the conscious mind does not perceive it as intentional movement. And when a patient with divided hemispheres is asked to explain a body movement it did not perceive, the patient gives a response that is logical and well reasoned... and usually confabulated. And when the answer is challenged, they produce another that is equally fabricated. The conscious mind is functioning in this case as a sports announcer, assigning thoughts and intents without any actual knowledge of either.
Enter the LLM. The LLM Chatbot does not know what it is saying. It does not understand the question in any human sense, nor does it have any stake in the answer; It generates a response to a query that follows as logically from the premise as its training model supports. That's it. The Chatbot is giving an answer that makes enough sense be accepted, based on millions of similar questions and answers.
Just like the people sitting around the Ouija board insist they are not moving the planchet, the consciousness of the players do not perceive their underlying contribution to its movement so they generate the next most believable answer: ghosts. Just like the patient with a divided hemisphere has no knowledge of what their physical movement was, so their consciousness just hallucinates a believable explanation.
The UI of an LLM is the consciousness of the machine. It is neither fully subservient to the LLM nor is it in control if it. But in all the ways that matter, it behaves exactly the way the human consciousness does when disconnected from unconscious cause and effect.
This morning I had a brief episode of what I have taken to calling "Being out of phase with yourself." It felt as if I was parroting my own actions as I was performing them, as if I was suddenly and briefly cognizant of the delay between my unconscious movements and my conscious awareness of them.
This particular incident led me once again ponder the age old question of Agency versus Determinism: Am I actually in control of my own actions? This led me down a rather fascinating rabbit hole reading work on derealization from Metzinger to Hume to Libet, and ultimately arriving at the traditional contradictory conclusion, Yes & No.
But one new thought did enter my mind that seemed like one worth following. Consider the ideomotor effect and the experiments of Gazzaniga and Sperry on divided hemispheres. The unconscious mind can and does control the body with zero input from the conscious mind, and not just in autonomic nervous system reactions; the hands move on the Ouija board and the conscious mind does not perceive it as intentional movement. And when a patient with divided hemispheres is asked to explain a body movement it did not perceive, the patient gives a response that is logical and well reasoned... and usually confabulated. And when the answer is challenged, they produce another that is equally fabricated. The conscious mind is functioning in this case as a sports announcer, assigning thoughts and intents without any actual knowledge of either.
Enter the LLM. The LLM Chatbot does not know what it is saying. It does not understand the question in any human sense, nor does it have any stake in the answer; It generates a response to a query that follows as logically from the premise as its training model supports. That's it. The Chatbot is giving an answer that makes enough sense be accepted, based on millions of similar questions and answers.
Just like the people sitting around the Ouija board insist they are not moving the planchet, the consciousness of the players do not perceive their underlying contribution to its movement so they generate the next most believable answer: ghosts. Just like the patient with a divided hemisphere has no knowledge of what their physical movement was, so their consciousness just hallucinates a believable explanation.
The UI of an LLM is the consciousness of the machine. It is neither fully subservient to the LLM nor is it in control if it. But in all the ways that matter, it behaves exactly the way the human consciousness does when disconnected from unconscious cause and effect.