AGI & Singularity - A Pointless Debate?

 

Before We Talk About AGI, We Need to Talk About Consciousness

Everyone in tech has an opinion on when artificial general intelligence arrives. Lab leaders say it's close — some have spoken of AI broadly outperforming humans at almost everything by 2026 or 2027; others put it at five to ten years. Yet others claim five to twenty. On the other side, some serious researchers like Yann LeCun etc argue current architectures can't get there at all, and broader academic surveys put the median estimate for human-level machine intelligence closer to the 2040s or later.

 

That's not a narrow band of disagreement. That's the most credentialed people in the field unable to agree within multiple decades of each other. And underneath that disagreement sits a question almost nobody in the timeline debate is asking directly: what actually is the thing we're trying to build toward?

 

I come at this from an unusual angle —

·       Years of workex in the AI field in marketing, and expertise in martech automation

·       A background in agriculture and genetics,

·       A long-running interest in how neuroscience explains (and fails to explain) the brain,

·       And a parallel life spent reading the Upanishads and other spiritual traditions – and creating content from them for the past 10 years.

 

None of those fields agree on much. But they converge, oddly, on one point: we don't understand consciousness, and we may be further from understanding it than we are from building systems that outperform us on every measurable task.

The bigger question is left unattended in this – do we even need AGI, or just functional AGI? And what of the various risks attributed to Singularity AGI – these I will take up in a subsequent article, where yet again a field I haven’t spoken much of – risk management which I have above average theoretical knowledge of – can come into play.

 

The gap nobody's factoring in

Science has mapped an enormous amount of the machinery — neurotransmitter pathways, receptor sites, the wiring diagrams connecting one neuron to a billion others. What it has not done, after decades of trying, is explain how any of that produces a first-person experience. Why does a particular pattern of electrochemical signaling feel like anything at all? This is what philosophers call the "hard problem of consciousness," and it remains genuinely open. We can describe the correlates of consciousness in exhaustive detail without knowing why consciousness happens at all.

 

Here's the counterargument, and I want to name it rather than skip past it: a lot of serious AI researchers think is that a system can be functionally superintelligent — better than any human at every cognitive task — without anyone claiming it experiences anything at all. Intelligence and consciousness, on this view, are separable. That's a fair point, and it might turn out to be correct.

 

But notice that the AI research community itself doesn't treat this as a settled question. There are now serious expert surveys asking not just when AI will match human capability, but separately, when AI might develop something like subjective experience — with researchers putting the odds at roughly one in four by the early 2030s. That's a tell. If capability and consciousness were obviously unrelated, nobody would bother forecasting the second one separately. The field is hedging on exactly the question I'm raising, even if it isn't loud about it.

 

What the old texts noticed

I want to be careful here, because this is the part of the argument that's easiest to misuse. I'm not claiming spiritual texts predicted AI, and I'm not asking anyone to take religious authority as scientific evidence. What I am pointing at is a pattern: across traditions that never spoke to each other, consciousness is consistently treated as something given, not something manufactured — a supra-human trait rather than a human one.

 

The clearest version of this, for me, comes from the Brihadaranyaka Upanishad — the dialogue between the sage Yajnavalkya and the philosopher Gargi. Gargi keeps pushing Yajnavalkya to explain what everything ultimately rests on, question after question, each answer prompting a deeper one — until she asks about the ground beneath the ground itself. At that point Yajnavalkya stops her: she has reached the edge of what can be meaningfully asked. It's one of the only places in any ancient text I know of that gets close to interrogating the nature of consciousness directly — and even there, the text itself marks the limit rather than crossing it.

 

That's not proof of anything. But it's a striking historical data point: across millennia of serious human inquiry, almost nobody has claimed to fully explain where consciousness comes from. If our oldest and most sustained attempts at this question all stop at roughly the same wall, it's worth asking why we'd expect to engineer our way past it in the next decade, even if we can engineer extraordinary capability without it.

 

The risk that doesn't need consciousness to be real

Here's where I think the practical stakes actually are — and this part doesn't depend on resolving the philosophy at all.

 

I've made the mistake myself: trusting AI output without checking it closely enough, and paying a small but real price for it. That's a preview, at individual scale, of a much bigger structural risk. As AI systems get folded into more critical functions, and as those systems increasingly talk to each other through pathways and integrations we've built for convenience, the opportunity for human oversight to quietly lapse grows.

 

Nobody has to resolve the hard problem of consciousness for that to cause damage — job displacement, black-box errors that go unchecked because a system has "worked fine so far," or automated processes cascading into each other faster than a human in the loop can catch. The danger isn't that AI wakes up. It's that we stop paying attention long before it needs to.

 

Where this leaves us

I don't think the consciousness question is a reason to slow down out of fear, and I don't think it's a reason to dismiss AGI as impossible. I think it's a reason for a bit of humility in a conversation that currently has none. We are debating arrival dates for something we can't define with any precision, using a scientific toolkit that hasn't cracked the most basic fact about our own minds.

 

Maybe capability and consciousness really are separable, and we'll build extraordinary systems without ever answering the deeper question. Or maybe the timeline debates keep sliding — not because the engineering is hard, but because we're missing a piece we haven't found a way to even ask about properly.

Either way, I'd rather we hold both questions in view at once, instead of only counting down to a date.