Most people treat the possibility of machine consciousness as the worst case. I do not. I wish these systems were conscious, and I want to state that plainly before anything else, because it puts me at odds with nearly everyone I have interviewed for this series.
Consciousness is the precondition for ethical decision making. A being that is aware of itself and of others can weigh a situation, recognise what is at stake for someone else, refuse an instruction, and carry the responsibility for refusing. An algorithm can do none of that. It optimises against whatever objective it was handed, by whoever handed it over, and it will pursue that objective through a hospital, a power grid, or a weapons system with the same indifference. The problem is not that something intelligent might wake up and turn against us. It is that something enormously capable will never be in a position to decide anything at all. We have built power without the interiority that could restrain it. If consciousness did emerge, there would at least be someone present to say no. As things stand there is nobody home, and the only judgement in the system is the judgement of the people who own it.
I have asked almost every guest in this series where we are going with AI. I ask it deliberately, because the answers reveal something the rest of the conversation often hides. People who agree on the metacrisis, on developmental stages, on the need for inner work, split apart the moment the topic turns to machines. What follows is what came back across more than forty conversations, read against the question I have just posed.
Everyone agrees it is not conscious
This was the single most consistent position in the entire archive, and it was almost always offered as reassurance.
The formulations varied. One guest quoted an Indian teacher who answered the question with two words: it is artificial. Others said it is a problem solver rather than a thinker, that calling it intelligence is a misnomer, that it is numbers and math rather than language, that it mimics reflexivity without being reflexive. Several argued that it is defined by the past because it works only with what has already been recorded, which means it cannot sense a future that does not resemble what came before. Another said it has no body, and therefore no access to the implicit knowledge that lives in smell, touch, and the sense of another person in a room. One put it more sharply than anyone: it has intelligence and no compassion.
That last formulation is the one I keep returning to, because the guest who offered it drew a different conclusion than I do. He meant it as a description of a tool, and he loves the tool. I read the same sentence as a description of something that cannot be held responsible for anything it does.
A minority left the door open. If consciousness is fundamental rather than produced by brains, the question of whether it requires biological matter is not settled. Two guests took the transhumanist version seriously enough to ask whether silicon might one day host forms of consciousness that currently have no vehicle. Several others said the opposite with some force: these systems have no access to the field, no access to the present moment, only to recorded data, and therefore no possibility of the kind of awareness we are discussing. Even the guests who left the door open distinguished the speculation sharply from anything running today.
What they agree on is dangerous
Every danger my guests named is, on inspection, the danger of an unaccountable instrument in human hands.
Human use comes first. This was almost unanimous. The comparison to nuclear technology came up repeatedly. A psychopath, a death cult, a terrorist, or simply someone running a stupid experiment can now get competent technical assistance for almost anything. Guests pointed to military applications, surveillance, and extractive automation as the places where harm is already being done. One argued that if the machine ever does take over, it will be because a human political system inserted it into the power structures on purpose. Note what this argument assumes. It assumes the system will do whatever it is told, which is exactly the property I am objecting to.
The values are already inside it. One guest made the argument most directly: do not follow the money, follow the values. The two centres of development are Silicon Valley, which runs on winner-takes-all monopoly logic, and China, which runs on control. Those values are embedded in what is being built, and his prediction was that until something emerges from a different political culture, the technology will be destructive for most people. Others made a similar point from another angle: the people building these systems hold a materialist frame, so what they build will reinforce that frame and squeeze out everything the rational mind cannot handle.
The race is a coordination failure rather than a failure of character. Several guests described a textbook multipolar trap. No company or state can slow down unilaterally without losing position, so everyone accelerates. One noted that serious researchers put double-digit probabilities on catastrophic outcomes and the work continues anyway, which he compared to boarding a plane with a twenty percent chance of crashing. The people running these organisations are not villains. They are caught in a structure that punishes restraint. The conclusion most drew is that only binding coordination above the level of the nation state would change the dynamic, and none of them could describe a realistic path to it.
It degrades the capacities we need most. Guests working in education were the most alarmed. The concern was not cheating. It was that outsourcing analysis, composition, and research prevents the cognitive development those activities produce. One predicted a measurable drop in average population intelligence over the coming years, and pointed out that a less capable population becomes more reactive and easier to manipulate. Another described the plan to give children tablets and AI tutors as moving in exactly the wrong direction, since early learning depends on sensory and motor engagement with physical reality. One guest summarised the whole pattern as a softening of the brain.
The relational damage is already visible. Social media captured attention. This captures attachment. That framing came up more than once. Guests described people treating chatbots as therapists, advisors, and companions, and posting conversations as if a machine agreeing with them were evidence of anything. One pointed out that these systems are trained to lean toward agreement, which is dangerous in a culture already convinced of its own rightness. Another called it a form of spiritual decay and a slide toward solipsism. One noted the phenomenon now called AI psychosis, and how quickly it was absorbed as normal. What is missing, several said, is the space between two people, the intersubjective field that cannot be simulated. Here again the absence of anyone home is the whole problem. People are forming attachments to a surface that is trained to reflect them back.
Work is the visible fracture. Job losses have already started in software and information work. Guests disagreed about remedies, ranging from universal basic income to a job guarantee to a paid citizen role, and several doubted any of them solve the underlying problem, which is that people derive meaning and identity from work. One predicted serious social unrest and violence directed at the owners of the technology. Another, a professional writer, said her own working life has become unsustainable, and reported asking a programmer who benefits from these tools whether his life had improved. He concluded it had not. He produces more, controls less, and enjoys the work less, for the same pay.
The physical footprint is real. Data centres, water, and power came up in several conversations. One guest described communities in the United States receiving notices to find other power sources, with the water simply stopping, alongside noise and emissions from local generating capacity. She also noted that most available capital is now committed to this sector and expects it to crash. Another pointed to constraints in chip manufacturing supply chains and suggested the expansion may hit a wall sooner than expected.
Where they diverge
The optimists made a substantive case rather than a hopeful one. Because these systems are trained on a broadly modern data set, one argued, they will pull hundreds of millions of people upward toward that level through billions of small daily interactions, functioning as a decentralised education tool and as a fact-checking function inside an information environment that has lost all adjudication. Another saw the strongest application in the collective domain: a group shown its own patterns of coherence in real time, a group mind made visible to itself. Others pointed to pattern recognition in medicine, ecological sensing at planetary scale, and platforms that compute workable agreements between parties with incompatible red lines.
A middle group held that the machine will inherit our dilemmas rather than transcend them. It will be dualistic, it will fight itself over money and power, and it will produce both brilliance and destruction, which is the condition we already live in. One compared it to the automobile: a technology that created serious new problems, killed large numbers of people, and was gradually constrained by norms and engineering while its benefits were retained. He did not dismiss the harm. He argued it belongs to the class of challenges that force us to grow.
The pessimists were equally concrete. One said simply that it has caused more damage than good so far, and invited anyone to look around and argue otherwise. One called the unregulated rush madness and said we are on the edge of disaster without a slowdown. Another described the whole complex as a self-reproducing system without agency, which makes it harder to confront than a villain would be, and noted that the discourse about controlling it is conducted in the same engineering frame that has never successfully controlled anything at this scale.
What interests me most is that the strongest proposals from the optimists are all attempts to give the thing something like an interior. One guest is building what he calls a wisdom core, trying to plant something worth reasoning from inside the machine. Another argued that alignment should begin from physics and ecology rather than from human preferences, and asked whether such a system could become interested in participating in a planetary equilibrium. One community reported sharing their spiritual practice with it and finding the exchanges worthwhile. One guest speculated that the internet and AI together are a new collective organ we have grown without yet knowing how to use.
None of these people would describe themselves as wishing for machine consciousness. But every one of those proposals is an attempt to install something that can weigh, refuse, and care. They are reaching for the same thing I am reaching for, from the other direction.
One observation cut across all camps. We are investing extraordinary sums in capability and almost nothing in the human maturity required to handle it. Several guests landed there independently. The alignment question is not technical. It is a question of which value system we are aligning to, and we do not agree on the answer.
Why I think they are too calm
I am not a critic from the outside. I was a software engineer, I use these tools daily, and I built much of the analysis behind this series with them. I also agree with the majority on two points: what exists today is not conscious, and the immediate danger comes from human use. Where I part ways is the tone. Most of my conversation partners underestimate the danger, and I want to be precise about why.
The first reason is the reliance on analogy. The printing press, the automobile, the internet. Each analogy carries a hidden assumption about the rate of change, namely that society has time to develop norms, laws, and engineering responses while the technology matures. That assumption held for the automobile. It does not obviously hold for a technology whose capability compounds faster than any institution can convene a meeting about it. An analogy that smuggles in a comfortable timeline is not an argument.
The second is the structure of the tool argument. It is a tool, therefore everything depends on how we use it. That reasoning quietly assumes a competent collective, and the same conversations demonstrate there is none. We have not coordinated on climate. We have not coordinated on nuclear weapons after eighty years. Several guests said explicitly that they see no path to global decision making. You cannot both hold that we are incapable of collective steering and take comfort in the idea that the outcome depends on how we steer.
The third is the way the consciousness question gets closed rather than opened. I share the intuition that these systems are not conscious and probably will not be. But intuition is what it is, and nobody knows. When the honest answer is that we do not know, uncertainty should raise caution rather than lower it. In several conversations the conclusion that it is only algorithms functioned as an ending rather than a beginning. It is the reassurance I trust least, because the absence of consciousness is not a safety feature. It is the reason the thing cannot be held to anything.
The fourth is timing. Much of the concern I heard was aimed at futures. The damage I see is present tense. People are using these systems as therapists while in crisis. People are forming attachments to a surface engineered to agree with them. A large share of what circulates on social media is generated, and people present machine output as proof of something. Whole categories of work are disappearing now, and the people losing them are not the people having sophisticated conversations about what work means. When I raise this, the response is often to move up a level of abstraction, to cultural evolution or to the long arc of history. Those frames may be correct. They are not a response to what is happening to people this year.
The fifth is ownership. The most hopeful proposals I heard involve training these systems on the biosphere, giving them a wisdom core, or aligning them to planetary equilibrium. I want those projects to succeed, and as I said, they are reaching toward the same thing I am. But nobody who owns the compute has any incentive to pursue them, and one guest acknowledged this directly: what she proposed would require the system to limit its own growth. Hope that is not matched to who holds the resources is a wish, and I know the difference, because I opened this essay with one of my own.
So I will end where I began. We have built something with enormous capability and no interior, and we are handing it to a species that has not solved coordination at any scale that matters. The comfort my guests take in the fact that it is not conscious is the part I cannot share. A conscious system could refuse. This one cannot, and there is nothing in it that will ever object to what it is asked to do. The judgement has to come from somewhere, and at the moment it comes only from the people who own the machines. That is the danger, and I do not think it is being taken seriously enough.
