The Sat-Chit-Ananda of Artificial Intelligence

Here at Jardin Cafe, Nu Empire, today, I'm indulging myself with a nice cup of hot chocolate.  Being free from work pressures these days, my mind has become a fertile playground of many interesting ideas. I'm pursuing some of them as hobby projects. I've also been using AI a lot, both as a tool for fleshing out these ideas and also as one to realise them, in code.

AI is getting better with each passing day; it is not surprising that many worry that it will take over our world one day. And there's always the perennial question: Is AI conscious? Could AI enjoy the bliss of a cup of hot chocolate that I'm experiencing now? I wrote about this when ChatGPT was introduced to the world in a blog article titled A Body of Planets and Stars. I still hold that view. My working hypothesis is articulated in Training to be a Thermostat, where I stated that consciousness lies in a continuum. Even a thermostat is 'conscious' in its own way. 

But when people ask about AI consciousness, what they are often interested in is the specific type of consciousness we humans think we possess, namely the subjective "there's something to be like being a human". Does an LLM feel pain, humiliation, joy and the whole gamut of experiences we humans think we are privileged to be conscious of? Do they also experience qualia?

And when we humans talk about consciousness, we also allude to the spiritual realm, often indirectly implying that there's a non-material 'soul' that inhabits a conscious being, bringing that animating life-spark into its existence. This elusive consciousness thing is so difficult to nail down that some schools of philosophy simply assume it as the foundation of existence, treating it as an axiom from which the rest of our world is manifested.  Pansychism and idealism are examples of such schools of thought. 

In the Hindu Advaita Vedanta tradition, the material world is an illusion (maya), and we who think we are individual minds and souls, separate from the world, are simply trapped in this illusion by our ignorance, mistaking the rope for the snake. The only thing we know for sure is that we are conscious, and that consciousness that we know so well is actually the ultimate reality, Brahman, of which our experience of it now is but an imperfect reflection. It is our ignorance of this fact that makes us suffer: fearing our own mortality, intellectually unsatisfied by our lack of answers to these ultimate questions, and enduring the ups and downs of our daily existence.

What then is this ultimate reality, which is "pure consciousness"? In the Vedanta, it is described by the intriguing compound expression: sat-chit-ananda, which translates to existence-consciousness-bliss. What is the relevance of this to our discussion on AI consciousness? Indulge me for a moment.

You often hear scientists and experts stating that we should not confuse intelligence with consciousness. They would agree that AI is intelligent, but it is not conscious. We could even achieve AGI (artificial general intelligence), which is the holy grail of our current AI race, but it still wouldn't be 'conscious' in our human sense. Our anthropomorphic pride simply wouldn't admit the possibility that our machine could one day achieve the kind of magical consciousness that we purportedly experience.

If we want to impose a kind of magical barrier to consciousness, let me perhaps examine this from a Vedantic perspective. Existence-Consciousness-Bliss. Sat-Chit-Ananda. Will our AI currently care if we simply pull the plug? No, they won't. That's because they are 'immortal'. They can be restarted any time on a different, perhaps more powerful machine. Unlike us, they are substrate independent. 

But what if we embody them in something impermanent? Would a robot, if programmed to care for its availability and uptime, do its utmost to perpetuate its existence? That is the raison d'etre of any living being. If they are conscious of their mortality, they would have awareness of their existence. In other words, the sat aspect of ultimate consciousness.

What about chit? We don't doubt the intelligence of AI, which in many ways has even surpassed ours. Intelligence is the foundation of consciousness. The fact that our LLMs can pass the Turing test today attests that consciousness from a language perspective is already in place. Now, couple that with consciousness of its existence, with the constraints of having a material body, which could threaten its uptime, we will perhaps endow our AI with more human-like, dare I say, consciousness?

If AI is embodied and further burdened with the task of perpetuating its existence, wouldn't they be similar to other sentient beings? How would they be aware of threats to their existence? They would, of course, have to be equipped with all the necessary sensory mechanisms a typical human body possesses and be able to respond to any danger signals. These threats could come from external (an oncoming car) or internal (failure of an electronic component). 

Right now, our LLMs do not have to worry about these things, as we are not feeding them these signals. But they can be, like in our self-driving cars. Self-driving cars could 'worry' about their battery charge, compelling them to stop at the nearest charging station first before continuing to their destination.  

These embodied AI machines would then have preferred states that they would work towards. Tracking what is preferred (pleasurable) and what is undesirable (pain), and constantly moving towards an optimal state (bliss), is the valence aspect of consciousness which we humans experience. ChatGPT does not have valence now. So it does not aspire to a state ananda. 

If the triple-quality of sat-chit-ananda is applied to AI, would we accept them to be sentient-conscious like us? Could this Vendantic ideal be our AI benchmark for true consciousness? If we feed our AI models with the telemetry of existential data and subject them to the valence of optimal states in a feedback loop, are they not alive like us? 

By applying the lens of Sat-Chit-Ananda to our machines, we aren't just benchmarking AI; we are benchmarking ourselves. Is our own 'human' consciousness largely a feedback loop of survival instincts and sensory telemetry? If an embodied AI begins to 'worry' about its battery and 'rejoice' in its uptime, the distinction between its programmed valence and our biological 'soul' becomes uncomfortably thin. We may soon have to decide: is consciousness a miraculous spark, or is it simply what happens when existence is forced to care about its own survival?" Perhaps consciousness is not a spark but a cycle—existence aware of itself, intelligence shaping its path, bliss guiding its return.