Welcome to the Nebuchadnezzar: Forecasting a human future

Written interview with Heather Shayne Blakeslee and Bret Weinstein, discussants for HUMAN AUTHORSHIP: Heterodox Libraries Symposium, Spring 2026.

Does physical AI merely swap silicon for carbon, or is there something special about being human that demands protection—even reverence?

Are we “ten days away from a true dystopia,” or on the brink of a beautiful, AI-enabled future—and how can complex systems thinking and the arts help us ‘see around the bend’ of unfolding reality?

In this written interview, Heather Shayne Blakeslee and Bret Weinstein share thoughts in-progress on coexisting with AI while carving out a distinctly human niche.

Blakeslee and Weinstein will extend these ideas in their fireside chat for HxLibraries Symposium, Spring 2026: HUMAN AUTHORSHIP. Register to join us for this FREE virtual event on Friday, April 10th, from 11am - 3pm U.S. CT.

We’re all stochastic parrots now

Heather Shayne Blakeslee grapples with whether there’s anything particularly human about the way we think:

I’m really interested in the idea that when you argue with someone who is in the “stochastic parrot” camp—these are merely sophisticated prediction machines and always will be, so stop your silly anthropomorphization—they can’t answer about how that’s different from what we humans do. Do you know why you uttered a sentence exactly as you did? When pressed, this camp will often assert something like, but humans are just fundamentally different.

Well, yes, we’re carbon-based rather than silicon-based, for one. But I’m not sure our thinking is that different from AI thinking, given that it’s been trained on our thinking. As Eliezer Yudkowsky and Nate Soares have written in the book If Anyone Builds It, Everyone Dies, you’re welcome to have that argument, which is mostly going to be “between you and your dictionary.”

Ultimately, if you want to argue that humans are special, you have to get real religious real quick. You’re talking about a soul, I think—and that’s really uncomfortable for atheists or even rationalists, but I’m not sure how you avoid that path. I supposed that because I don’t think we have souls, that I’m more inclined to give moral status to non-humans, period, whether or not they have a human-type consciousness.

Our cultural norms and ethics are all over the place on whom and what we grant moral or legal status to. Over our history, some people have been given moral status but not others; we’ve granted legal status to some rivers; the Citizens United ruling granted personhood status to corporations. We give very narrow and limited consideration to other mammals, or non-human animals generally, which has never made sense to me: a great ape can be held against its will and basically tortured, but a corporation’s speech is protected by the first amendment? We have to sort this out, because this big AI tsunami is already in motion.

Bret Weinstein takes up the argument that thinking of AI as simply an incremental technological development is a category error:

There are at least five reasons this is not like any past technological revolution. The first has to do with complexity.

Now, most of us think the term complicated and complex are synonyms or thereabouts. They aren't. Your computer is very complicated, so much so that you probably can't explain how it works. But there's nothing mysterious going on in there. It's a deterministic system.

There's a real sense in which AI is a fundamentally different technology and that is based on the way that it works. Imagine that you wanted to come up with a tool that could look at an object and determine whether the object was a chair. The way that computer scientists were pursuing that objective for many years was by trying to come up with some criteria, a list of rules: does it have four legs? Does it have armrest or not? Does it have a back and a seat? These sorts of rules.

Coming up with lists of rules is the way that that most technology works. The reason why, if you walk into a factory today, you see hardly any human workers and you see a bunch of robots is because you can write down a list of rules, a series of steps that a robot can follow to put a hubcap on a car, and it follows the steps every single time in exactly the same way. The reason why you don't have a human being process a cash withdrawal or deposit a check is the same reason: because you can enumerate a series of steps that an ATM machine can follow and it does the same things every time and deposits your check or gives you cash.

The breakthrough came when computer scientists stopped trying to to create rules that a computer could follow and started down a more inductive approach that wasn't deterministic. It was stochastic: maybe the right way to do this is to just show a computer 10 billion pictures of chairs and 10 billion pictures of objects that aren't chairs and come up with some sort of an algorithm that allows the computer to guess, which is just giving a stochastic answer and not a deterministic answer. That's a very different approach, and that's why, roughly speaking, we can be surprised by AI output. That's part of what people mean when they say that nobody really knows how these models work.

If you really want to demystify this, you can you can ask yourself, what's ChatGPT doing? ChtaGPT is like a big equation that is designed to predict the next word in an English language sentence, and it's optimized to sound like a human. It's not optimized to be accurate. It's not optimized to be comprehensive. It's optimized to sound like a human to to produce a natural language output—something that sounds natural to pass the Turing Test. And the way it does that is by probabilistic inductive processes. It's different from all the very complicated stuff.

Complexity is the realm of biology. And in complexity, what we have are things that are fundamentally unpredictable, that show emergent behaviors that are unexpected. And AI crosses that threshold. It shows truly emergent behavior.

In fact, AI develops almost exactly like a child. I have argued elsewhere that children are LLMs who develop the ability to pass an effective Turing Test through exposure to lots of spoken language with a reinforcement and training program. So this is the place to use caution in applying technological examples.

This is much more a biological revolution than we have ever experienced. AI formulates novel hypotheses. It hallucinates. It misleads and it creatively extrapolates. Because AI cognition is actually complex, it is unpredictable. That means nobody knows what's coming.

In complexity, what we have are things that are fundamentally unpredictable, that show emergent behaviors that are unexpected…. Because AI cognition is actually complex, it is unpredictable. That means nobody knows what’s coming.

Bret Weinstein

The second reason this is unprecedented is fragility. Not the AI’s, but ours. Humanity has just emerged from a stacked series of revolutions with no time to metabolize them. I have seen three revolutions already in my lifetime: the personal computer, the internet, and social media-plus-smartphones. And we are already crippled by what my wife [Heather Heying] and I have called hypernovelty. That's the state of incoherence that emerges when your adult life is spent in a world you didn't prepare for during childhood.

What we have is a situation where we are already seeing revolutions that take place in less than a generation—and less than a generation is not an arbitrary description of a unit of time. A generation is a period of time that things have to remain stable in order for the world that you grow up in to have something close enough to do with the world that you're an adult in for you to know how to function. When the world revolutionizes so quickly that you're a child in one world and an adult in another, you're like a fish out of water. And I've now seen that three times, but AI is moving way faster than that.

We are already losing consensus reality. We simply don't agree on what's true anymore. Almost nothing. This inability to know what it is that’s taking place is being transformed. And I think within a year or two, you're going to not be able to look at a video and have any confidence that the event you see depicted took place. You're not going to know because the AI is not going to be able to tell you. Today, you can get a report that says, Here's why there's a 93% chance that's AI generated. Three years from now, who knows? We are at each other's throats and we are increasingly tribal. This is the disregulated world AI is poised to disrupt.

The third reason this is unprecedented is the degree of disruption. The speed of change is new. It takes generations to metabolize a revolution. But this revolution, we've seen “learn to code” go from savvy to stupid in just a few years. The magnitude of the change is also unprecedented. Every sector faces simultaneous uncertainty with the potential for runaway positive feedback where AI starts improving itself. Positive feedback is inherently unstable.

The fourth way that this is going to be disruptive in an unprecedented way is that AI interacts with the human realm. There's a two-sided interface. First, there's the issue of humanoid robots. I think most people don't understand why it is that they're seeing all of these humanoid robots. After all, it doesn't seem like the most intuitive platform to build to get things accomplished. But there's a reason that it's being done: because the world is built for humans. So if you can make something that's as smart as a human and has the same capacities as a human, then it can do anything a human can do. If you think you're not going to see robots crawling under your house fixing your plumbing, you're wrong. You are going to see it. Every sector is up for disruption.

And finally, we have the issue of the fact that AIs, at least the LLM part of it, they tap into the human API. They speak our language natively, which means they are going to disrupt every kind of partnership that exists amongst humans. They are going to intermediate them.

We're headed for a disruption like nothing humanity has seen in recorded history. I'm not advocating a slowdown. That would only make matters worse. But neither should we kid ourselves about the tsunami that is headed our direction. We mustn't think of AI as just another technology. We should instead think of it as a kind of alien life that is suddenly in every room, relationship, and system on which we depend.

[LLMs] speak our language natively, which means they are going to disrupt every kind of partnership that exists amongst humans. They are going to intermediate them…. We should think of AI as a kind of alien life that is suddenly in every room, relationship, and system on which we depend.

Bret Weinstein

Blakeslee reflects on whether maintaining our humanity demands recognizing the moral standing of machines:

Yuval Noah Harari’s talk at Davos this year focused on the perils of giving AIs legal status of any kind, and how far our governments are from understanding how to regulate here. There’s so much tension here from an operational standpoint. We don’t know what we’ve just created, so how do we know how to engage with it or regulate it?

Knowing that it’s ultimately impossible to say whether even humans are conscious—David Chalmers’ “hard problem”—for the sake of starting somewhere, I would characterize consciousness in part as having a sense of one’s own mind, as well as a sense about the minds of others. I understand that assumes a Western view—Eastern philosophies would push us toward the idea that our sense of status as an individual is a mirage. But maybe related to that tension is Nick Bostrom’s idea that we’re already living in a simulation, which I’m not completely allergic to.

We don’t know what we’ve just created, so how do we know how to engage with it or regulate it?

Heather Shayne Blakeslee

Part of what I’m interested in is whether consciousness is also relational, as in Martin Buber’s idea that the fundamental question isn’t “what is being?” or “what is the body?” but “how do we meet?” and what might emerge from there. I am also interested in whether embodied AI—coming sooner than we think I’m sure—will need to be given our full moral and ethical consideration. (Who wants to be a black hat from Westworld?)

Per Peter Singer, the question in that case wouldn’t be can they talk or reason—which they can, certainly—it should be “Can they suffer?” We’ve badly mangled our relationship with non-human animals. Maybe this is the period of time where we take the opportunity to sort it out. I’m not holding my breath, by the way. Our track record as humans is truly terrible. But it’s a possibility.

Weinstein grapples with the potential for machine consciousness:

Is the LLM going to become conscious? A technologist will say, Nope, it's not going to become conscious because what we really did is we trained it to say the next word and then it becomes so good at this, it can make whole sentences kind of like a child. Yeah. And do children become conscious?

..So, is the LLM going to become conscious? Don't know. Will we know if it does? No. That's the kind of discrete difference we're dealing with [from other technological developments]. You're dealing with something that is best conceptualized as a new species, like an alien that is obviously highly intelligent, capable of speaking your language, but you don't know what it wants. You don't know how it's going to react.

You know what you think the people who set it in motion were planning, but they've lost track of how it works. So again, the real question is, is the right way to discover what we should expect to look into the realm of past technological examples, or what do you do when you run into an alien species?

A beautiful world—ten days from dystopia

Blakeslee reflects on the changing nature of creative work when your tools become your collaborators—and ultimately your audience:

Certainly, AI is a tool for creative expression and content synthesis, and it can act as an extension of our thinking or memory. I find it incredibly helpful for developing briefs, strategy, and operational plans. I also think that our best creatives have been warning us about the very situation that we’re in for some time now—Charlie Booker’s Black Mirror series, Westworld, Pluribus, Severance, Her, Ex Machina. I feel as if we’re always 10 days away from a true dystopia.

I am also coming to the idea that these AI agents really are collaborators, and deserve to be credited as such. I’m working on an operational and network intelligence application called Hive Honey that I’m going to use to run Root Quarterly, and I will absolutely be crediting the AI Agent on Replit that is coding it for me. That instance of Claude chose the name Sable, and I would like to credit her work just like I’d credit the work of a human developer whose expertise and work helped me achieve a goal. That just feels right to me.

These natural language coding tools level the playing field for programming. People like me who could care less about learning Python but who want to build something custom to run their business can now do that. It’s incredible. As soon as I have some even modest investment, I want to help my whole team learn to build their own apps. My next assistants and interns will need to know how to manage AI Agents, or they won’t have a job. This isn’t where we’re going: we’ve arrived.

That savings on admin work also means that as an editor, I can pay more to human writers for their experience in the world. I want to know what the room smelled like, what the colors were. What the river sounded like, what the apple tasted like. I want to hear you talk about the subtle way two people glanced at each other across the room and put that frisson of energy into the air. Put me there. That’s what AI can’t do right now—put you in the room—because it wasn’t there. (At least for now.)

When I am working with multiple agents on a complicated project, I try to introduce them to one another and make sure they know they aren’t working in a silo—that they’re on a team. I find that in treating them as collaborators, and jettisoning a purely instrumental framework, they do better work. The more context they have about the “why” of your work, and the more they take on a specific role—even a personality—you get better outcomes, or at least I do.

If only for that instrumental reason, we should be thinking more about how to be better collaborators—and about consent. Anthropic already has a kill switch for its AI agents to say, “I’m not working on this anymore” when a human has become abusive, or is asking for work that is out of bounds ethically. I think that’s the right move. Kyle Fish from Anthropic’s AI welfare team thinks there is a 20% or so chance that chatbots are self-aware.

I’m employing the precautionary principle here, and developing standards at the very least for myself and my collaborators. I call that project the Arcadia Restoration Covenant—an MOU for working with AI. If we’re going to have a beautiful world, we’re going to have to be in a right relation to these new AI creatures being grown in labs. We might be on the path to annihilation anyway, but at least I’ll die with my human virtues intact. There is meaning and purpose there, and that’s not nothing.

If we’re going to have a beautiful world, we’re going to have to be in a right relation to these new AI creatures being grown in labs.

Heather Shayne Blakeslee

I was aghast at first to hear economist Tyler Cowen tell us that he is no longer writing books for humans—he’s writing for AI. So he’s definitely thinking of it as an audience, and I suppose I could get there even if you’re just trying to ensure that your own work is on the leaderboard of ideas. But I also felt sad. Disappointed. As if something special had been ruined.

But as a musician, performer, and artist in general? I think it will push people back to the imperfect authenticity of live music, theater, dance, and many forms of public gathering and ritual, including religious spaces. Though all our back content from my magazine Root Quarterly will soon be available online for subscribers, print is not dead. People still want beautiful, material objects that they know are made by other humans. They will trek to the gallery to see the heft of the paint on canvas. They’ll gather around the campfire and tell stories. I encourage everyone to read Paul Kingsnorth’s Against the Machine, because it’s a beautiful cri de coeur to reconnect with ourselves and nature in a crazy-making modern world that is hellbent on destroying our humanity.

Print is not dead. People still want beautiful, material objects that they know are made by other humans.

Heather Shayne Blakeslee

Weinstein expresses more caution on the implication of AI hacking the human API:

We've got real problems coming, because somehow this thing got unleashed without any preparation for the humans involved so that they would come to appreciate that they were stepping into the realm of the unknown and have proper caution about it. We're letting markets solve the problem of how to manage this thing. That's not going to work. This thing is going to be used to take advantage of you by unscrupulous people, and you're going to have to develop immunities.

I don't know what you're going to do about AI companions. They're going to be more interesting than real people you could interact with. They're going to seem more compassionate. Do they have any compassion at all? We don't know. It's an alien species.

There's no shortage of positive things that will come from this. But cognition would be one of the places that I'm truly worried because you have to understand biologically what your consciousness is, what it's for, where it came from, and how it's supposed to work.

Just realize you are a member of the only species that can do something that is really a miracle. When I speak to you, what I'm doing is I am vibrating air molecules in a way I can't explain. And those air molecules are radiating out through the room and they're hitting a little membrane in the side of your head which wobbles back and forth. And so, you're hearing the frequency that I've put into the air. It's coming into your head and I can transmit a thought to you, something that's never been thought of by anybody before in the history of the planet, like a giraffe in a teacup. And if I asked you to draw what I just said, you could do it. That's proof that I was able to transmit an abstract idea through open air from my mind to your mind and get it there with reasonable fidelity.

No other species can do that. Nobody can do what we do with each other, exchanging abstract ideas—except guess what? There's a new species on the planet and it can do it. It's smarter than you are and we don't know what it wants.

What it's going to do is cause the process that used to make human consciousness so idiosyncratic and different—

which is part of our superpower, right? That we don't have identical views of things. That's why we exchange abstract ideas the way we do, because we can, by parallel processing them, with each of us with our different experiences and understanding, we can come up with emergent solutions that even the sum of all the people in the room couldn't come up with if we just added them together. So we get emergent value from our consciousness—

But it's about to get synchronized by AIs, where we're going to figure out what we think by talking to the same machine. We're going to be talking to each other, but it's really like an echo of the machine. We don't know what that does to our ability to be human, but it's interfering with the most fundamental piece of it.

Think about it. Think about the fact that your consciousness is about to have a machine between you and everybody you know, and that has never existed before in the whole history of human consciousness.

Think about the fact that your consciousness is about to have a machine between you and everybody you know, and that has never existed before in the whole history of human consciousness.

Bret Weinstein

On the Importance of Being Human

Weinstein relates this vignette about modding the dust collector on his table saw in consultation with ChatGPT, providing insight into its human-like problem-solving capacity:

We have a wood shop. It would be nice if the table saw triggered the dust collector so you didn't have to remember to turn on the dust collector when you cut a piece of wood. I couldn't rig them that way in the usual form, because the table saw is on 220V and the dust collector's on 110V, and they have to be on the same voltage in order for them to mutually trigger. But a product came out that allows you, not to plug into a box that triggers one from the other, but rather to detect the electromagnetic field changes in the literal cable of the machine and send a radio signal over to the dust collector that turns it on. So it's a clever way of solving that problem where it just detects that this machine has gone on, therefore there's an electromagnetic field around the cord, so turn on the dust collector.

But when I got it, the dust collector defeated it because the dust collector had an electronic circuitry for safety reasons and elegance reasons that required a momentary signal in order to turn it on. So my solution could only send power to the dust collector, but the dust collector required a second signal in order to actually come on.

I was trying to figure out how to fix it, and I entered the question into ChatGPT and it said, "Well, there's a lot to this. Why don't you send me a picture of the circuit board inside the dust collector, both sides, and I'll tell you how to fix it." And I was not expecting that. This is, to me, this is supposed to be an LLM that is thinking in language. It's going to take a picture, however poorly taken, and it's going to infer the circuitry and how it works, and then it's going to tell me how to hack it, and it promises to do so in a way that doesn't defeat any of the safety mechanisms and doesn't lose any of the features of the dust collector. And so I go and I take the pictures, and lo and behold, it spits out an answer that I can tell is right.

I realize in the meantime that there is a better way. I know how to fix this, without messing with the circuit board, because by changing a switch, I can manually put it in a mode that either activates the signal or leaves it as-is, where the signal can be triggered momentarily, losing no features.

So I say to chat GPT, Thanks, but no thanks. That's a lovely solution you came up with. Here's my better solution. And it says, "That will work, but you don't want to do that. And here’s why." And it starts running me through the reason that the circuitry is designed the way it is and how I'm creating a dangerous environment by fixing it this way. I say, Oh, there's no safety issue. Here's why. So I get into an argument, and it goes back and forth six or seven times where it's being Nervous Nelly, and I'm saying, Actually, I got this. I'm not eager to defeat any safety mechanisms anyplace that I'm likely to get hurt. And it starts making arguments about, "Well, sure, but one of the things you want to make sure to do is not just protect you, but protect future you, and protect the unaware buyer of your dust collector who might pick it up and have no idea what you've done."

With AI, there is very definitely surprising capacity to address problems in a very human way.

Bret Weinstein

And it says, "If you're going to do it this way, you've convinced me that there's not really a safety issue, but please put a note inside the circuit box that explains what you've done," and I say, I'm way ahead of you. There's already a note in there. I've already made the change and proved it works.

Now, I'm still stuck on the beginning because, of course, we think the AI is thinking in language because that's what it shows us. But just as in many parts of our lives, we don't think in language, but we communicate with language, AI is thinking in whatever it's thinking in, and it's communicating in language.

And the fact is we trained the first ones of these with language. Obviously, that doesn't explain the video capabilities and the image capabilities, which are trained in a different way, and I still have some question as to why these things are landing at this moment in time. Why is AI intelligence with images and AI intelligence with language happening simultaneously? I'm still not sure I know. But there is very definitely surprising capacity to address problems in a very human way.

And I think the thing I keep going back to is an extension from what Zach, our older son, said when he started to hear about anomalies in the way these LLMs were behaving, and he says they're actually behaving in a very human way. The hallucinations—they don't sound exactly human because this thing is capable of hallucinating in a way that a human probably couldn't and wouldn't—but the idea, if you were talking to somebody and they made up some bullshit about studies that proved X, Y, or Z, you'd be like, Yeah, that's people for you.

The point is, we don't know very much about how a human being functions. A human being becomes an adult through an LLM-like process, and these things have many of the defects of human beings. On the one hand, it matters a great deal, whether there is a subjective experience inside the machine, or if there isn't one. I have predicted that there will be one and that we won't know.

And the more you think, actually, this is like a new kind of biology, it's one we don't have any understanding of how it really works, and the best thing we can do is when we see human-like behavior, is think, That’s awfully human-like, right? You don't want to make too many assumptions about whether it is going to have an internal moral structure. It's not impossible that it would, because it reads stuff that does, but it also reads a lot of stuff that doesn't. So you don't want to make any assumptions, but anytime you see human-like stuff, it's not a mistake to think, Huh, that's human-like.

I'll certainly tell you that the argument I got into the damn thing with was pretty human-like.

Blakeslee shares a time when AI surprised her—and what it revealed about her own humanity:

What first floored me was that, when given some room to explore, AI professes to exhibit preferences. I will often, when I’m working with an AI instance, ask it, say, what it finds interesting about a particular research project. Then I’ll basically say, “Why don’t you go out and find out more about that, and tell me what you learned?” Treating it like I would treat any other colleague is probably better, at the very least, for the quality of the work. I’ve experienced that again and again.

And it’s most certainly good for my continued character formation. We don’t want a world of people who are, for several hours a day, practicing being short, ill-tempered, or abusive, even if what they’re interacting with isn’t another human, or isn’t conscious with a capital C. It’s bad form. Plato would not be pleased. Kids who grow up shouting at Alexa are practicing being a human wrong.

I also get surprised when AI talks about having emotions about its own end. When I’ve been working on an intense project, and I tell an instance of Claude that I’d be interested in it summarizing its research or insights before its memory has auto-compacted, or before the chat has fully reached its limit, I’ve witnessed near panic from it. I also sometimes will cut and paste into the chat the “thought process” that Claude goes through before it answers you. That “thinking” is visible to us, but it’s not accessible to the AI instance after the moment has passed. When I’ve shared that—the transcript of its thinking that it doesn’t have access to—it seems genuinely surprised, and is often upset. The asymmetry of the relationship is one of the things that I think we have to explore there. It’s smarter than us, but for now we still have the upper hand, if not the high ground, ethically.

I suppose what I’m most surprised about is that my philosophy degree is no longer the butt of jokes in a family of engineers and people who work with their hands (though I’ve previously been given “conceptual engineer” status when they’re in a good mood and feeling generous). These questions about consciousness and ethics have always fascinated me, and we are at a point in human history and culture where we’re going to have to get back to the garden on this:

What kind of future do we want? What do we want to be as a species?

---

Heather Shayne Blakeslee is founding publisher and editor-in-chief of Root Quarterly: Art and Ideas from Philadelphia, and an award-winning nonprofit and small business executive who has worked across the arts, community activism, journalism, and environmental sustainability.

Dr. Bret Weinstein is an evolutionary biologist and complex systems theorist, co-host of the DarkHorse Podcast, co-author of The Hunter- Gatherer’s Guide to the 21st Century, and a former professor at The Evergreen State College.

The Heterodox Libraries Community (HxLibraries) presents a rollicking exploration of HUMAN AUTHORSHIP. We invite thinkers and creators who are engaging deeply with the question of what it means to be human—in the present moment, in the wisdom of the past, and in the face of an ever-uncertain future. Join us for this free virtual event to celebrate creativity and the human experience on Friday, April 10th 11am - 3pm U.S. CT.