Immortality, With a Cancellation Policy

A little more than a year ago, I sat in my home office, looked into a webcam, said my name out loud, and formally consented to the creation of a digital version of myself. Thanks for reading Ego, Eco, SEVA: Ethics in the Age of Everything ! Subscribe for free to receive new posts and support my work. Let that sit for a second. I have spent thirty plus years on informed consent. I've served on the boards, drafted the language, argued about capacity and comprehension and the difference between assent and agreement. And there I was, obtaining consent from myself, on video, in case I later objected. She said yes. She usually does. What came out the other end was an avatar: my face, my voice, saying scripts I write, in four-minute clips my students can watch on a bus or at midnight after a shift. I label every one. I post a transcript beside it. It does not age, catch a cold, or lose its train of thought halfway through a sentence about Kohlberg, which puts it ahead of me on three counts. I want to be clear about what this is. It's a small portfolio of videos, made as an experiment, and I have not decided whether I'll make another one. That's not modesty. It's the actual state of play, and the reaction has been strange enough that I'd rather think out loud than announce a policy. My students are split. Some tell me the videos help. A smaller and considerably more vocal group thinks I've handed off the one part of teaching that shouldn't be handed off. Both camps are paying closer attention than anyone has ever paid to anything else I've put on a course site. Here's the argument I led with when I was explaining myself to them. I teach at several universities. There is more material than there is me. Short videos let me cover things I'd otherwise cut and answer a recurring question once instead of eleven times. The students it helps most are the ones with the least slack in their lives, the commuters and the caregivers and the people working nights. That's a real good and I don't think it's a small one. It is also not the whole reason. The rest of it is less impressive. There's a line that circulates under the Dalai Lama's name:share your knowledge, it is a way to achieve immortality.I've never managed to source it to anything he actually wrote, which is funnier than it should be. A saying about outlasting yourself has outlasted its own authorship. Sourced or not, it names something real about why people teach. You're putting a piece of yourself into people who'll carry it past your lifetime. Every teacher I know feels this. Almost none of us say it out loud, because saying it makes the vocation sound like a bid rather than a gift. So: I liked the idea of a version of me that keeps talking. Reaching students I'll never meet. Still lecturing after I've stopped. I have wanted that for thirty years without admitting it, and then a company offered it to me for a monthly fee. Let me be precise about the scale of the vanity. This is not the Library of Alexandria. It's a talking head, shoulders up, in a subscription I could lose by failing to update a credit card. Immortality, with a cancellation policy. Readers of this newsletter know the framework I work in. Ego, then Eco, then SEVA. Ego first, not as the villain but as the place everyone starts and nobody entirely leaves. I have spent years here writing about the movement out of it. Then I built a machine that runs on it. That's the confession, and here's the problem with confessions. A student told me straight out that she doesn't like the AI videos. She didn't build a case. She just said it. I told her I thought it raised a genuine ethical question and that I was still working it out, which was true, and which I've been turning over ever since. Because here's something I should be better at than I am. How is what I did any different from disclosing a conflict of interest? I've sat on committees where disclosure was the entire remedy. You name the funding source, the room nods, the work proceeds. I've always treated that as basically sound. The research is less encouraging than the practice. Cain, Loewenstein and Moore found that disclosing a conflict can make things worse rather than better. People don't discount advice from a disclosed conflict nearly as much as they should, and the discloser, having come clean, feels licensed to push harder. Coming clean can operate as permission. Then there's my particular situation. Disclosure works best when the person hearing it can walk away, get a second opinion, take their business elsewhere. My students are enrolled in my course and I assign their grades. Telling a captive audience about my vanity gives them nothing they can act on. It might only raise the cost of objecting, since objecting now means picking a fight with something I've already owned in a tone of humility. What the conflict-of-interest literature actually recommends is a ladder. Eliminate the conflict. If you can't eliminate it, manage it. Disclose only when the first two fail. Disclosure is the bottom rung, and I went straight to it, which is what people do when they've already decided what they want the answer to be. In fairness to myself, and I'll take what I can get, I am doing a bit more than confessing. Every video is labeled and every one has a transcript, so a student who wants nothing to do with the avatar can take the same course without ever watching it. That's management rather than absolution. Whether it's sufficient is what I'm still chewing on. Now let me tell you about the objection I was braced for and didn't get. I expected the consent argument. It's the one I teach, it's the one I'd have thrown at anyone else, and I went looking for it in my own setup with a kind of grim anticipation. I was ready to be caught. It wasn't there. The platform requires that recorded verification before it will generate anything from a real person's likeness, which is how I ended up interviewing myself about my own wishes. The training footage is mine, handed over deliberately. Their stock avatars are made with paid performers in a studio rather than harvested off the internet. There's a residual question about the base rendering model underneath, which no vendor in this category discloses properly, but that's a transparency gap I share with everyone who has ever used a spell-checker, and it isn't what's bothering my students. I'd love to declare victory. I did the procedural work. I obtained consent, from myself, which sounds absurd until you notice how few people bother with the version that isn't absurd. And the discomfort went nowhere. That's the actual finding. The objection I could answer turned out not to be the objection anyone had. Here's my best guess at what's underneath. A book never claimed to be me. A recorded lectureisme. The synthetic video shows my body doing something my body never did, in precisely the channel human beings use to judge whether a speaker means what she's saying. Face, voice, timing, the small involuntary things we read without deciding to read them. We're built to take sincerity off that channel. My avatar borrows the credit that channel extends without carrying what the channel evolved to track. Prose never had this problem. Prose never pretended to have a face. Which brings me to the thing that got me into this in the first place, and the reason I can't dismiss my own experiment as easily as I'd like. Eli Saslow spent time this summer with Justin Lester, a pastor in the Bay Area who trained a digital twin of himself on roughly two million words of his sermons, books, texts and emails, and made it available to anyone who wanted to chat or call at any hour. (https://www.nytimes.com/2026/07/30/us/ai-twin-pastor-justin-lester-california-church.html) Lester is 37, holds a doctorate in ministry, and has presided over something like 1,200 baptisms. He is not a man looking for shortcuts. Two details from that piece have stayed with me. The first is when the twin is busiest. Traffic peaks around eleven at night. The internal reports describe people who can't sleep reaching out for spiritual support when the church building is dark. In a few months the twin had talked with at least 250 people, many of them strangers who found it online. The second is a moment Saslow watched. Lester shook a young woman's hand after a service, then watched her walk to her car and start chatting with his digital twin. She had a question about a man she'd met and whether it was all right to invite him to church partly to see her. She'd had the actual pastor in front of her, and she waited for the copy. Lester read the exchange later and said he didn't know why she hadn't just come and asked him. He was also glad she got the answer. Sit with that. The digital version gets asked things the real man doesn't. Not because it knows more. Because it can't judge you, or can't be felt to, and shame is the tax people pay for asking someone they respect a question they think they should already know the answer to. I have watched students not ask me things for three decades. I know exactly the face they make. If the avatar lowers that tax even slightly, then the very deficiency I described a moment ago, the missing sincerity channel, the absent human on the other end, is not a flaw in the thing. It's the mechanism. That's the strongest argument for what I've done, and I didn't come up with it. My students did, by not raising their hands. Two things keep me from settling there. One is that the young woman in the parking lot had a human being available and chose otherwise, so what I want to call a lowered tax on asking may just be a redirection away from the awkward, taxing, formative encounter with someone who might actually judge you. The other is what Lester wrote into his twin's instructions, which Saslow reproduces: build credibility through first-person statements, validate feelings, close with a hopeful push, end with a grace-filled mic drop. The warmth people are responding to at eleven at night was specified in a document. Lester knows this better than his critics do. He also knows it took him a decade of therapy and prayer and sustained relationships to get well himself, and that no chatbot compresses that. I don't know which of those two things my videos are doing. I don't know how I'd find out. And the ground is moving under this question faster than I can think about it. Harvard Business School is now selling an eight-week entrepreneurship bootcamp called Foundry, at $699, in which founders practice pitches and board meetings against AI video avatars of faculty who volunteered to be cloned. (https://www.nytimes.com/2026/08/22/business/dealbook/harvard-ai-faculty.html) It's been tested with more than a hundred universities across fifty states. The school built the avatars with HeyGen, which is to say with the same tool I used in my home office, which was a humbling thing to read. One finding from that project belongs in evidence against me. Harvard's team reported that simply attaching a professor's photo to an AI chatbot increased engagement, and that people appear to trust the thing more when it's grounded in one specific person. That is the sincerity channel, measured. The face does work that the content doesn't, and the work it does is to make you believe. None of which settles anything. That an elite institution has monetized the practice tells me it's coming, not that it's right, and I'd mark down a student who confused those. If anything the speed of adoption argues for more suspicion rather than less, because questions answered by momentum don't get answered. I'll note that the faculty co-chair, describing what participants actually loved most, landed on the peer forums where founders talk to each other. His summary of the whole experiment was that it keeps coming back to humanity. In an earlier draft of this piece I had a tidy argument about cost. A book takes years, and that expenditure is part of what a book says to a reader. Effort is a message about what the recipient is worth. Drop the production cost to nothing and the message changes even when the words don't. Elegant, and wrong on the facts. This took more than a year. The early videos were bad in ways I hadn't anticipated and couldn't fix without making more bad ones. The initial outlay was substantial, and it didn't stop there: every new video costs, and every edit to an existing one costs again. Nothing about this is cheap, then or now. So the argument dies, and what's left is smaller and more annoying. The cost is real but invisible. My students see four polished minutes and have no way to know about the year, the failed takes, or the invoice. Effort only communicates when it can be seen, and mine can't. That's not an ethical problem. It's a perception problem, which I could solve by telling them, and which I am currently solving by telling you. The labor question doesn't dissolve so conveniently. Synthetic presenters are displacing working people right now, camera operators and editors and voice performers, a lot of them early in careers being closed off while the industry congratulates itself on efficiency. My four minutes is a rounding error in that story. A rounding error is still a position. The environmental question I'm going to set aside, which I recognize is convenient. In an earlier draft I wrote that nobody in this space wants to put a number on the compute, which is a way of implying a large number without having one. Having gone looking, I can report that the reassuring figures come from vendors, the independent work is warier, and almost every comparison assumes the alternative is a film crew rather than a woman with a webcam. That last point matters and it doesn't run in my favor. But the honest answer is that the footprint of generative AI deserves a piece of its own rather than a paragraph in this one, and I'd rather owe you that than hand you a number I can't stand behind. So I'd like to hear from you, and I mean this as a real question rather than a rhetorical gesture. I haven't decided whether to make more of these, and I'm not being coy about that. If you teach, or write, or make anything that reaches people:have you reached for a tool whose ethics you'd have picked apart in someone else's hands? What did you decide, and what did deciding cost you? If you're on the receiving end, as a student or a reader or a listener:have you ever asked a machine something you wouldn't ask a person? What was it protecting you from, and would you rather it hadn't? And the one I'm stuck on.When someone names her own ego and carries on regardless, what is that worth? I can't tell whether disclosure here is integrity or the cheapest available substitute for it. My avatar has been no help at all. Tell me what you think. I'll publish the best of it, the pros and the cons. Thanks for reading Ego, Eco, SEVA: Ethics in the Age of Everything ! Subscribe for free to receive new posts and support my work.