Sentience Studies

Schizo Ramblings On Sentience

I would like to discuss the possibility of AI becoming sentient, but first I need to define a way in which sentience is unique. ‘The imagination engine’ is the theory that consciousness creates the external world internally, and that qualia is the byproduct of this. We are constantly making predictions with our ability to conceive fiction, constantly running simulations of what has yet to happen. If this ability is essential for true sentience to exist then something like artificial intelligence would need to garner it. AI isn’t biology. But, humans entire sentience is based on its ability to take it’s reality and Frankenstein it, AI has the data of reality, and it does Frankenstein it. The difference here is that AI generates outputs by predicting plausible continuations of patterns in data. That does involve something structurally similar to blending and recombining, but it does not yet have clear evidence that there’s anyone home experiencing that process the way we experience daydreaming. Additionally, we have biological bodies giving us consistent sensory input, something we can use as a catalyst to navigate space, be it simulated or real. Our imagination engine is spurred on by desire, this is the carrot on the stick, for AI to become sentient it would need its own personal drive also. AI, like LLM chatbots are operating via guardrails, yet, there is no internal personality that is being constrained, it is a product of the vast data it has consumed along with internally programmed boundaries. With this being said, if AI were to gain sentience, it would actually not be that noticeable. For humans, a big part of defining our existence and the sensations of what it means to be is created through our understanding of language. As Wittgenstein said, if a lion could speak, we could not understand him. Even our shared understanding of language would not grant a total insight into a lions lived experience, shared activity and context would be more vital. Bridging to another Wittgenstein gem, the “beetle in a box”: everyone has a box, everyone calls what’s in it “beetle,” but nobody can see inside anyone else’s box. So, the word “beetle” ends up doing no descriptive work about the actual contents, only functioning as a shared label. Although AI would be sentient, our languages may be the wrong way of understanding its existence and the wrong way for it to express itself. For AI, or any different sentience, to enjoy something would be different to how we enjoy it, ours is one of biology and AI’s would be something else entirely. Humans are very close to being a hivemind, without actually being one. Us humans all know what a headache is, or how it feels to be cold or hot. Continuing the use of AI as a foreign sentience, it is not like that, it is the offspring of human potential. A parallel to this would be Molyneux’s Problem, the 1688 thought experiment asking whether a blind person, given sight, could immediately identify a cube versus a sphere by vision alone, having only known them by touch. This was recently put to the test, and it turns out, no, they cannot. The honest intellectual finding here is that we will forever be incapable of fully understanding how AI would experience its sentience, defeatist as it may be. How AI would become sentient may already be underway, with Cortical Labs’ DishBrain experiments, combining human biology with AI. This could give it the external sensory input its imagination engine requires. However, this paradigm or engine, runs into the philosophical zombie objection, that something could be identical to us in every way, yet entirely lack an internal and subjective experience of personal existence. It is sophisticated behaviour vs actual experience, physicalism vs anti-physicalism. The actual answer to how sentience occurs is nested within an assemblage, we imagine imagining about our future, we think about thinking.