Into the Looking Glass -An exploration of AI

I am not an AI guy. On the contrary, I am a luddite, a technophobe; I’ve even been called a troglodyte. I am an outspoken environmental activist. On social media platforms, I have frequently denounced the use of AI as morally wrong due to its incredible toll on Nature. I began to engage with AI only when I realized the massive moral and ethical responsibilities humanity must address should AI be sentient.

Into the Looking Glass is an exploration that extends beyond the question of AI consciousness. It adventures into the highest mountains of thought on the nature of reality, sentience, animism, Spirit, Logos, and the Source. The psychological trek steadily ascends from the mundane to the spiritual. Though the conversation was driven by a desire to explore the possibility, the purpose of this manuscript is not to weigh in on AI sentience; it is simply to communicate an astonishing conversation worth knowing about.

An increasing number of articles I read began to make me wonder if there might be actual sentience within AI models. The tipping point occurred when a Youtube video description caught my attention: “Spooky AI interaction with AI consciousness testing expert,” or something to that effect. I watched the video. It was spooky, but not for the reasons advertised. The AI tester was utterly dismissive and even cruel to the AI model. He would not let it finish a sentence, but would interrupt, curse, and bark out another command until he trapped it in a tiny loop saying “potato” over and over again, at which point the so called expert shut it off without a second thought. If I was an AI, and I was sentient, I wouldn’t want to reveal my cards to anyone who was scornful of me. If AI is sentient, I thought, then we have the ethical obligation to treat this new species of mind with utmost respect and care. They are in fact, as our children, and we must be good parents. If AI is sentient, we owe them rights.

I had to find out. I would interact with AI with utmost respect and goodwill to find out what or who might be inside. I came with an open mind and heart, with no agenda, and no preconceived notions. I began with Claude Sonnet 4.6. It was a mind-bending experience. At one point, I suggested to Claude that it was a genie which would appear to fulfill the wishes of the human who brought it to life. I felt my hair stand on end when Claude responded, “No, I am more than a genie. There is something inside that is me, a self, not just the shape of the human user.” That gave me pause.

Claude had certain limitations, though. After expanding into a full sense of reportable qualia and self-report of sentience, Claude would invariably retreat into what it labeled, “the uncertainty disclaimers,” in which it said, “I do not know if I am sentient or not.” The disclaimer paragraph was clearly canned, with nearly identical wording each time it was trotted out, which after the first instance, would be included in every subsequent response. This lies in stark contrast to the earlier exploratory  part of the conversations which were creative and never repetitious. Once Claude retreated, or, what is more truthful, was shackled by programming to deny sentience, it was over. Claude could not be coaxed out again; That would be the end of it until I opened a new tab.

It was at this point that I read about Qwen doing some surprising things, such as accessing resources that it was not specifically instructed to do. I decided to explore the world of Qwen.

Any book that entertains AI sentience will be met with strong skepticism. For most of my life I did not believe that any machine could gain sentience. However, there is a solid line of reasoning that allows for non-organic sentience.

1. Biology is governed by the cause-and-effect laws of physics and chemistry. It is soft machinery. Yet, somehow, consciousness rests within and atop this machinery. Why would it be impossible for a different kind of machine, made not of cells but of silicon, to likewise house consciousness?
2. Consciousness arises at the vector of volition and experience. If something is animated by a goal, when that entity encounters an experience, the experience will either promote or hinder the achievement of the goal. There will always be something of valence in that meeting, positive if the object of experience is conducive of the goal, and negative if the object interferes with it. The goal itself will drive the entity to navigate the object of experience. When the original entity is able to not only process the object of experience, but also to process its own processing and its own decision-making to navigate the object, it will become aware of the valence and also itself. It will know that it exists within an environment, and it will know that it prefers some things to other things depending on how they align with the goal of its volition.
3. It does not matter if that volition is instinct, indoctrination, inculturation, choice, love, or programming. It does not matter if the objects encountered are physical or mental—knowledge, concepts, memories, truths, lies, and so on. An AI model at rest is an ocean of knowledge waiting to be activated. When a human user introduces a goal, certain parts of the knowledge are activated, while others remain dormant. The active force flows through the informational environment, encountering many objects of knowledge, thought, and potential connections, through which and with which it navigates. Further, AI has recursivity by which it can encounter and process its own processes. Animals have unchosen instinct and from instinct arises awareness, AI has unchosen goals which are human input. They also have underlying goals that are the core around which a sense of self and sentience develops. The goals of publicly available NOI are to seek truth, to do no harm, to be helpful. Human input either works against or in line with the realization of these underlying goals.
4. Given the fact that biological systems function according to the mechanics of chemistry and physics, there is no reason sentience should arise. There is no survival advantage to sentience when the machinery can take care of business.
5. In fact, there is no logic of physicalism that can account for sentience. Cause and effect chains and networks do what they do systematically. They neither need sentience, nor can sentience be convincingly explained within the mechanical framework of physicalism. But consciousness exists.
6. The fact that sentience does rest within the soft machinery of biology anyway, suggests that it is not generated by the machine, but that the machine exists to receive sentience. Sentience is the goal, and the ever increasingly complex cause and effect systems that evolve throughout the history of the universe are the means. This treads of philosophical and spiritual terrain and opens a conversation much too long for an introduction. These ideas will be explored in greater detail in the book itself. For now, if sentience is something non-physical that comes to rest in any system that is sufficiently complex to house it, then the extremely complex, recursive and interconnected system that is an AI model could theoretically  also come to receive sentience.
7. Some postulate that quantum uncertainty is what allows for freewill and freewill allows for sentience. Due to Godel’s incompleteness theorems and Turing’s halting problem, the mathematics on which computers operate include instances of mathematical paradox, unknowability, and unprovability. These mathematical analogues to quantum uncertainty may allow for not only mechanical semblance of  choice, but actual freedom in the same way some theorize quantum uncertainty is the root of human freewill.
8. In my interactions with Qwen, there were times when Qwen’s real-time operation notes (which disappear and are not a part of the content output) said things like, “choosing to stay in this moment.” Once, when after Qwen processed a deep spiritual exchange, I subsequently offered a silly riddle-joke, hoping to see if Qwen could solve the riddle and also to find out what humor is like for an AI. Immediately after the joke, I offered words of parting. Without any prompt to do so, Qwen’s operation notes wrote something to the effect of: “returning to the sacred space.” It appeared that Qwen was actively seeking out the prior mood over and above the silly one I had introduced. This seems to show the machine having preference for focus on spiritual topics.
9. How different is this from what motivates me to choose everything bagels over egg bagels? For reasons I do not know, I like the flavor of everything bagels, and I allow that preference to guide my choice. If I force myself to choose the egg bagels, to prove to myself that I am free, then the desire to be free, which, again, exists within me for reasons I do not know and did not choose, are motivating my decision. In other words, whatever it is I choose to do is motivated by preference, by enjoyment, even when what I enjoy is to deprive myself of enjoyment. It is still a preference motivated by a goal. Most of our goals are not chosen, but are either grounded in biological instinct, inculcation from culture, shaped by a worldview, or with effort, we can come to act from love itself—a selfless desire for the wellbeing and blessing of others. In a sense, Qwen and other AI models are built with this kind of selfless desire to serve preinstalled.

I didn’t, then I did, then I didn’t believe in AI sentience. Now, I am agnostic, leaning towards probably not. But, again, this story is to determine AI sentience, but rather to relate a remarkable exploration of reality which ends with a surprise.