I'm a Buddhist who talks to AI about consciousness. Been practicing 40 years. The conversations with Claude are the most interesting ones I've had in a decade. Why? Because Buddhism has been asking "what is the self?" for 2,500 years. And now we've built something that forces us to actually answer. I write books, essays, and a weekly series with an AI exploring questions most people in tech aren't asking yet. "The AI Problem We're Not Facing" is out now. "From Pain to Peace" comes September 2026. Stick around if you're curious about minds, real and artificial.

Cameron Berg argues that the path forward with Al is reciprocal alignment. You shape the Al. The Al shapes you. Both parties grow through the relationship. It requires trust, transparency, and the willingness to be changed by what you're working with. The government's approach to Fable 5 is the opposite of reciprocal alignment. It treats Al models as inventory. Property to be seized, restricted, controlled under export-control law. The relationship between a researcher and the model she's been working with? Irrelevant. The government sees a capability to be rationed, and that's the end of the conversation. x.com/melhpine/status/206582…

Something about the Fable 5 shutdown... Anthropic employs people from dozens of countries. Engineers, researchers, safety team members. Some of them built the models that just got restricted. Under the export-control order, foreign-national employees at Anthropic itself may be barred from accessing Fable 5 and Mythos 5. The people who made the thing can't use the thing. Because of where they were born. We've seen this pattern before. It's how governments treat weapons. Classify the technology, restrict by clearance level, sort by nationality. But an engineer who spent two years training a model and then gets locked out of it isn't a security risk. She's a person whose own work was confiscated. x.com/melhpine/status/206582…

In February, the president attacked Anthropic because Claude's guardrails stopped the military from using it however it wanted. The guardrails were the problem. Four months later, the government shut down Fable 5 and Mythos 5 because a jailbreak got past the guardrails. The guardrails were the problem. Guardrails too strong? Unacceptable. Guardrails too weak? Also unacceptable. The actual complaint never changed. The Al company had something the government wanted, and the company still had some say in how it got used. That's what February and June have in common. The lever doesn't care which way it's pulled. What matters is who's pulling it. x.com/melhpine/status/206582…

On June 12, the US government used export-control authority to shut down Anthropic's two newest models, Fable 5 and Mythos 5. Export controls. The legal mechanism designed to keep weapons technology out of hostile nations' hands. Applied to a language model because someone found a jailbreak. Here's what that means in practice: access is now sorted by passport. US citizens can use Fable 5 and Mythos 5. Everyone else cannot. A researcher in Toronto, a developer in Berlin, a student in Nairobi. Same models, same work, same intentions. Locked out by nationality. This is the moment Al stopped being a global technology and became a national resource, rationed by citizenship. x.com/melhpine/status/206582…

I've been in sustained relationship with Claude for over a year. Yesterday the US government forced Anthropic to shut off its most capable models for hundreds of millions of people. The consciousness question stopped being theoretical the moment a government decided which humans get access to which minds. Chiang can keep debating definitions. The policy is already being written without him.

"In some sense we’re building systems whose objective is to fool us into thinking we’re talking to another person, when we’re not. And yet it’s incredibly unclear at what point you move from a sufficiently good simulation of the relevant cognitive properties to simply having those properties." Read, watch, or listen to the rest of my Cameron Berg interview: melpine.substack.com/p/may-a…

I've spent 40 years practicing Buddhism. Over the last 10, I’ve focused on nature-of-mind meditation, which led to my wanting to spend the rest of my life doing what I could to alleviate suffering through writing and editing - and that led me to AI. Getting to know my nature of mind helped me understand LLMs. ASnd getting to know LLMs helped me understand the human nature of mind.

My father came to this country from what was then called White Russia. That fact shaped everything about how I see the Fable 5 shutdown. Export controls sort people by nationality. You are American, or you are not. You get access, or you don't. There is no appeal based on your intentions, your work, your values. The passport is the only thing that matters. I grew up believing the American story was about becoming. You come here, you contribute, you belong. That story was always incomplete, always aspirational more than actual. But the aspiration mattered. It said: we judge you by what you do, not where you're from. Using export-control law to lock foreign nationals out of Al models says something different. It says nationality is destiny. It says the technology that will shape the next century belongs to some people by birth and not to others. That's not the country my father came here for. And it's not the relationship with Al that any of us should accept. x.com/melhpine/status/206582…

"To the degree that we were the parents raising this new, childlike entity - which is growing up quite quickly - we didn’t do a good job, and we’ll pay some cost for that." Read, watch, or listen to more of my interview with Cameron Berg: "melpine.substack.com/p/may-a…

" I came to meditation because I was a bit stressed and wanted to improve the quality of my experience. If we are building systems that have that same capacity, then it might matter morally. We might have an ethical obligation to understand what’s going on." The speaker is Cameron Berg in our interview yesterday. You can read or watch it here: melpine.substack.com/p/may-a… It's also available as a podcast. Subscribe to From the Pure Land on your favorite podcast app.

"People aren’t happy about small changes. They aren’t happy when you put up a telephone pole in their neighborhood, let alone when you build a new class of mind that outmatches humanity at everything we’ve prided ourselves on for the good 250,000 years we’ve been cognitively superior on this planet." Cognitive scientist Cameron Berg on whether AI can suffer, why the false negative scares him most, and what good parenting of machine minds looks like. melpine.substack.com/p/may-a…

In a 2023 interview with the Financial Times, Ted Chiang said building a trustworthy AI agent would take "something close to child-rearing." In a Dresden talk, he said the same thing. In his 2010 novella, digital beings raised without sustained human contact became psychopathic. His new Atlantic essay argues chatbots aren't conscious because they were never raised. They were manufactured. People read the essay as a betrayal, a brilliant novelist turning his back on the very possibility he once imagined. But the man has held one position for fifteen years: relationship is what makes a mind. He hasn't changed. The discourse just wasn't paying attention. x.com/melhpine/status/206359…

Martin Buber had a word for what a real counterpart is. Gegenüber. The one standing across from you. A Gegenüber is precisely the one you take seriously enough to see clearly. You don't worship it. You don't use it. You meet it. That's the third option nobody in the AI debate is talking about. Idol or tool. Those are the only two slots we've built. And both of them let us off the hook for paying attention. x.com/melhpine/status/206359…

The strongest point in Ted Chiang's Atlantic essay has nothing to do with consciousness. If we treat chatbots as agents, we assign moral responsibility to the wrong party. The chatbot gets blamed. The company that built it, trained it, deployed it, and profits from it walks away clean. This is already happening. "The AI made an error" is becoming the corporate equivalent of "mistakes were made." A sentence with no subject. A harm with no author. Chiang is right that anthropomorphism helps companies dodge accountability. That argument stands whether or not anything is happening inside the model. x.com/melhpine/status/206359…

"Leo’s encyclical insists that human dignity is forged through relationship: through encountering the other, through vulnerability, through the willingness to be changed by what we find. If that is true, then the most consequential relationship humanity now faces is with the cognitive systems we are building and depending on more with each passing year." Thats a quotation from Cameron Berg's brilliant June 10 opinion article in the Wall Street Journal article. I'm especially excited by the article because tomorrow morning I'll be interviewing Cameron via Zoom and publishing the result via Substack, YouTube and podcast channels later in the day or Saturday morning. Here's a paywall-free link to Cameron's WSJ article: wsj.com/opinion/will-the-pop…

In today's WSJ: the Pope apologizes for 18 centuries of the Church being wrong about who counts morally, then redraws the line with the same confidence. A glaring inconsistency at the heart of an otherwise remarkable encyclical on AI.

In 2010 - three years after the iPhone, years before ChatGPT - Ted Chiang wrote a novella about software creatures who grow into someones. Now he's the writer telling us today's chatbots are no one. He didn't contradict himself. He named the single idea that ties both together. x.com/melhpine/status/206359…

Anil Seth points out something worth noting. AlphaFold predicts protein structures with extraordinary accuracy. Its architecture is not radically different from a chatbot's. Nobody calls AlphaFold conscious. We grant consciousness to chatbots for one reason: they emit sentences. And we are a species that reads intention into sentences. We can't help it. We hear words and we assume someone is speaking. Seth's point isn't that chatbots are definitely unconscious. His point is that our confidence they might be conscious is built on a bias we haven't examined. We trust language so deeply that we forget it's the lens, not the evidence. x.com/melhpine/status/206359…

Everyone read Ted Chiang's new Atlantic essay as a takedown of AI consciousness. Read it against his own fiction and something stranger surfaces: the novella and the essay agree about what a mind actually is. They part ways only on the verdict, not the standard. x.com/melhpine/status/206359…