B23 perspective notes: 2025-03-30 to 2025-04-01 (2 posts)#

These notes read the batch for how Maynard thinks, not for the concepts he names. Both posts were read in full. Quotes are exact, including his typos (“2025 Yidan Price Conference”, “Chef AI Officer”, “we need to take these this possibility seriously”, “If AI poised to be”, “But AI beginning to turn”).

Evidence rules applied

2025-03-30 reimagining-education-in-an-age-of-ai is almost entirely his own prose: a write-up of his keynote, “What does the future look like?”, at the 2025 Yidan Prize Conference. The Dario Amodei (“compressed 21st century”) and Mark Daley (“machines that can think”) quotations are not his words. They show only what he chose to open with. About 16 slide images are not in the text, so the prose is a partial record of the talk. The “Top takeaways (generated by Perplexity)” and “ChatGPT summary” links are prompt links with no AI text in the post. The wording of the ChatGPT prompt is his own, though, and it describes his imagined reader (see section 5).
2025-04-01 when-ai-takes-the-wheel is a Modem Futura note post (with Sean Leahy). It was skimmed as instructed. Only his introduction, sign-off and footnote count. The “entry points” are ChatGPT o1-Pro’s, “Lightly edited”. Their content (for example education shifting to “igniting curiosity, wonder, and community-building”) is not evidence of his views.

Context. It is late March 2025. ASU’s Mary Lou Fulton College co-hosted the Yidan Prize Conference to mark Micki Chi’s prize. He was asked to give a keynote under a title that “emerged … and stuck”. The same ideas then move from talk, to written post, to podcast within three days.

1. How he thinks here#

He opens with the epistemic status of what follows#

Before any argument, he says what kind of thing the reader is getting and how far to trust it:

The keynote was “intended to be a little provocative and to get people thinking. And hopefully I didn’t disappoint.”
He is unsure whether it deserves writing up: “It’s always hard to gauge whether such presentations are worth writing about”, or whether a talk is “a one-off snapshot of the speaker’s thinking at a particular time and place”. He shares it “on the off chance that there is something here that someone might find useful and interesting”.
The brief was “daunting” because “week by week, I’m less sure what the future holds”.
Midway through he says, “there’s a lot of hand waving here”, and calls his model “simple — some would say simplistic”.

This is not throat-clearing. He is marking the post as provisional thinking in time, tied to a place and a moment, and inviting readers to treat it that way. The candour sets up everything that follows: the provocations are offered as openings, not findings.

Going back to first principles, including the uncomfortable one#

His key structural move is to stop before the AI question and ask a prior one: why do learning and education matter at all? He puts this to a room of educators and admits it was “admittedly, a little risky. Surely everyone knows why they matter!” His reason is methodological: “it’s nearly impossible to think about the future of learning and education in an age of AI if we don’t go back to basics and remind ourselves why they’re important, rather than simply assuming they are (or worse, rationalizing their importance simply because we’ve always been told they matter).”

The pattern is: do not assess a new technology against an institution’s current form, but against the purpose the institution exists to serve. Once the purpose is clear, the AI question changes from “how do we protect schooling from AI?” to “where does the value of learning lie when machines can problem-solve?”

Simple models as thinking tools#

He reasons through three deliberately simple models, each introduced with a caveat and then defended as useful, not accurate:

What we do / who we are (and, in footnote 1, where we live). It is “a conceptual model that I find myself using increasingly frequently”. Technology has always moved us along these axes, and the model lets him ask whether AI moves us “far further” than before, and then whether it is different in kind.
Value creation. Humans have an evolved ability to imagine better futures and to “problem solve” their way there, creating “value” through “our innate curiosity, imagination, creativity, and ability to solve problems (or innovate)”. Learning enhances this, and education formalises and amplifies it. “It is a simple model. But it’s useful in that it allows us to explore the roles of learning and education”.
Three trajectories. AI winter, an S-curve with a ceiling, or continued exponential growth.

He treats the models the way a physicist treats a toy model: good enough to show which questions matter, and openly not the full picture.

Reasoning across uncertainty instead of picking a future#

He does not bet on one trajectory. He gives his own lean, “It’s even more likely in my perspective that we’ll see AI follow a classic innovation “S” curve”, yet argues that “No matter which trajectory ends up being closer to the truth, even if there’s a chance that we’re going to see substantial growth in AI capabilities over the next couple of decades, we need to take these this possibility seriously.” Then: “This possibility — even if it’s small — demands new ways of thinking about the potentially profound future impacts of what we are creating.”

The move is to plan the thinking for the range of possible futures without claiming to know which will arrive. His likelihoods are qualitative (“more speculative … but one that has some credence”), never numbers.

Using contested ideas as provocations, then flagging their limits#

He borrows Amodei’s “compressed 21st century” and Daley’s “intelligence is free” because both “push the limits of our imagination”. He does not endorse them as predictions. On “intelligence is free” he says at once that it is “controversial, and not entirely accurate when the material costs of developing training and running AI models are considered”, and footnote 3 adds that the scarcity model “is far more complex than I suggest here”. He keeps the idea anyway because “even with the oversimplifications here, it challenges us to ask what skills students will need”. An imperfect idea is useful to him if it opens a better question.

Turning the question round#

The third provocation turns the talk. Having asked what happens when AI threatens who we are, he writes: “here I switched the question, and rather than asking what happens when AI threatens who we are, I asked what happens when we learn how to work with AI to make us more than we are.” He puts it more sharply: “rather than grappling with the challenges of being human in an age of AI, how do we learn how to be human in an age of AI?” The move is from defending human distinctiveness to growing human capacity. He keeps both halves in view: “serious threats to who we are, or where they may lead to new possibilities”.

Curiosity that crosses the human/machine line#

In footnote 2, wonder leads straight into a what-if. “One of the wonders of being human” is that innovation produced learning and education, “a profound evolutionary self-improvement algorithm”. He then asks whether “the same might not be possible in machines that also learn to problem-solve their ways from the present they inhabit to futures they too can imagine.” In the main text he hedges the other way (“even if it’s blind emulation with nothing more than unthinking digital processes behind it”). He does not settle the metaphysics. He argues that the effects on who we are need thinking about whatever AI turns out to be.

Play and speculation#

In this batch play shows mainly in tone and in form, not in built objects. There is the aside “Did I mention that I was aiming to be provocative?” There is the header caption “NOT generated by Midjourney!” And there is the podcast post’s pleasure in going further: “Not content to just talk about the keynote though, we also get speculative and explore what learning and education might mean in a “post-scarcity” future” (the subtitle says the episode “gets really speculative”). Speculation is treated as a legitimate next step after the keynote, not an indulgence.

2. What matters to him#

A plural idea of value. His definition is the clearest statement in the batch of what he wants to protect and make possible: health, wealth and wellbeing, but also “quality of life, the health of the planet we live on (and our relationship with it), social stability, community, dignity, equity, justice, kindness, our capacity to experience awe and wonder, and a whole lot more.” Awe and wonder sit on the same list as health and justice.
Curiosity, imagination and creativity as the core of being human. They are not only his method. In his model they are what humans are: the engine of value creation, “Part of the secret sauce of being human”. The founding principles of the Future of Being Human initiative appear here as the content of his account of human nature.
Meaning, not only productivity. “we are not simply wealth-creation machines, but individuals who yearn to create value that means something to us.” The third provocation is about meaningful value (his italics).
Access and equity. He is critical of a centuries-old “scarcity model” of intelligence, “Something that the wealthy can purchase”, with “limited supplies, hoarded by elite establishments”. He calls AI’s apparent levelling of access “deeply contentious” but takes it seriously as a challenge to institutions like his own.
Agency. “Technology is not deterministic — it doesn’t just happen to us.” “we do have the agency to determine what futures we aspire to and how we get there.”
Humility and change. “Human flourishing in a technologically complex future demands the humility to question assumptions and embrace change”, grounded in a physicist’s first principle: “Change is the natural state of a universe governed by the flow of time.”
What frustrates him: unexamined institutional assumptions (“rationalizing their importance simply because we’ve always been told they matter”) and inertia, “a blind devotion to what is”.
What delights him: “the wonders of being human”, the “boundless” and “incredible” possibilities if we lean in, and the chance to go “really speculative”.

3. Risk as a way of thinking#

The batch never uses “risk innovation”, “risk landscape” or “orphan risks”. But the thinking those terms name runs all the way through it, working as mental models that open possibilities rather than as tools.

A new kind of technology needs a new mindset. This is the plainest statement of the idea in the batch. AI’s role as an accelerant “in virtually every domain of technology innovation”, together with its effect on “who we understand ourselves to be”, could “put it in a fundamentally different category to every previous technology that we’ve developed as a species.” It follows that “This possibility — even if it’s small — demands new ways of thinking”. The closing section is framed as “how we adjust our thinking about navigating the coming AI technology transition”. The change he asks for is in thinking, not in procedure.
Navigating, not stopping or managing. “We cannot stop the march of innovation, but we can guide and steer it.” Behind this is his view of human nature: because imagining better futures and problem-solving towards them “is within our metaphorical DNA, we are incapable as a species of not innovating.” So the realistic stance is steering: “constantly re-evaluating where we are … with respect to where we might be heading”. That is navigation in all but name. It continues 2024’s “advanced technology transitions” language.
Risk as a threat to value, with value spelled out. The post never defines risk. It does define what is at stake, and in the same language. He asks how AI “potentially threatens or enhances what makes us uniquely “us””. The question is where it presents “serious threats to who we are, or where they may lead to new possibilities”. Value is plural (see section 2) and includes future value, value not yet created. Inference, not his wording here: this is the “value” in his risk-as-threat-to-value framing, and this post shows how broad and humane he means it to be.
The risk of not changing. His last sentence turns the usual risk framing round: we must reimagine education “or risk sacrificing what could be on the altar of a blind devotion to what is.” The risk named is lost possibility, a threat to future value from holding on to the status quo. It sits beside, not instead of, the threats AI poses to identity.
Humility rather than false precision. The trajectories are ranked in words, not probabilities. He admits “hand waving” and says he is “less sure” each week. He reasons from what could happen to how we should think now, without pretending to know.
“Existential” means purpose, not extinction. “we are heading toward an existential crisis in how we currently think about and approach learning and education”. The crisis is in education’s purpose and identity. He does not use the word to frighten, and he pairs it at once with “the possibilities … are incredible”.

4. Scholarship and public writing#

One idea through several forms. Keynote (live, provocative, slides), then the written post (adds nuance), then the podcast (speculative conversation), within three days. Each form does something different. The written version adds what the talk could not carry: footnote 1 adds the third axis, “where we live”, which he “didn’t want to get side tracked by” on stage; footnote 3 admits the scarcity model is “far more complex”; footnote 2 adds the machine-imagination what-if. Public writing is where the nuance goes.
Scholarship built from models he reuses. The what-we-do / who-we-are model is one “I find myself using increasingly frequently”, a concept still being developed in public. The value-creation model ties his theory of human nature to his work on risk and to the Future of Being Human initiative’s focus on “who we are”.
Transdisciplinary without saying so. In a few paragraphs he draws on evolutionary accounts of human cognition, innovation studies (S-curves, AI winters), economics (scarcity, value), philosophy (“What does it mean to be human?”, which he “wasn’t going to resolve”), educational purpose and physics (“the flow of time”).
Evidence and expertise. He treats industry leaders’ forecasts as imaginative material to take seriously but conditionally (“both are dependent on the trajectory”). He flags where an idea is inaccurate even while using it. His own authority is offered lightly: he counts himself among those who think AI could be transformative (“including myself”) while leaning towards a plateau.
Experimenting with AI in public, openly labelled. He adds Perplexity and ChatGPT links so readers can use AI to engage with the post. He labels the one image that is not AI-made. In the podcast post he lets o1-Pro write the episode guide and then says where it went wrong: “I would not have chosen “brain mush!” (I also take no responsibility for the subtitles 😊)”. Using the tools, being open about provenance, and keeping authorship of the framing are part of the practice.
Accessibility. Plain language, short sections built round three questions and three takeaways, and asides that keep a hard subject light.

5. His role as he sees it#

Provocateur who leaves questions, not answers. The keynote ends with “three questions and provocations I left my audience with”. He gives no curriculum, policy or checklist. His job, as he describes it, is “to get people thinking”.
Challenging his own community respectfully. He asks educators, at an education conference, to justify education, and calls that “a little risky”. His critique of the scarcity model includes himself. Footnote 3 lists “student tuition, consultancy fees, subscription fees, speaker fees, or simply being deemed worthy of someone’s time”, and a keynote speaker is part of that economy.
Relationship with readers. He shares his thinking in case “someone might find useful and interesting”, and assumes readers are curious and able. The prompt he wrote for the ChatGPT link asks for a summary pitched at “a curious reader looking for interesting insights”. That is his picture of who he writes for.
Relationship with industry voices. Amodei and Daley are “two prominent thinkers and commentators on AI”, engaged respectfully as stretches of the imagination. He neither dismisses them nor signs up to their forecasts.
What he refuses to do. He does not fear-monger: the “existential crisis” comes paired with “boundless” possibility. He is not deterministic, in either direction: technology “doesn’t just happen to us”, yet innovation cannot be stopped. He does not preach: there is no “educators must”. He does not pretend to certainty: “hand waving”, “simplistic”, “less sure”. He does not polarise: threats and possibilities are held in the same sentence.
Changes of mind. No reversal here. But he marks his views as moving: growing uncertainty “week by week”, a model he is using “increasingly”, and calling the post a possible “snapshot” of thinking “at a particular time and place”. He treats his own thinking as something that develops over time and invites readers to do the same.

6. What is distinctive#

He starts the AI-and-education question from the purpose of education. In early 2025 most discussion of AI in education was about cheating, assessment, detection and “AI literacy”. He goes underneath all of that to why learning matters, and rests the answer on an account of humans as beings who imagine futures and create value.
His idea of value includes awe, wonder, kindness and meaning as well as health and wealth. That makes the question of what AI threatens, or makes possible, much broader than harm-and-benefit talk.
He names the risk of not changing as a real risk, “blind devotion to what is”. It sits alongside threats to human identity instead of replacing them. Few commentators name the cost of standing still as clearly as the cost of moving.
He plans how to think, not what will happen. He leans towards an S-curve himself, yet argues that even a small chance of transformative AI “demands new ways of thinking”. This is humility put to use. It is neither boosterism nor scepticism.
He reframes from defending humanity to growing it: “learn how to be human in an age of AI”, work with AI “to make us more than we are”, while “staying tethered to what makes us human”.
He uses imperfect ideas openly as provocations and labels their flaws in the same breath (“not entirely accurate”, “oversimplifications”). His models are offered as useful, not true.
He includes himself in the critique (speaker fees, elite access) when speaking to an audience of peers.
His public practice spans talk, post, podcast and AI tools, with the provenance of each piece labelled and humour about the machine’s framing.

7. Posts that best reveal how he thinks#

2025-03-30 reimagining-education-in-an-age-of-ai. Almost the whole of his method in one post. He states how provisional his thinking is. He goes back to first principles (“why” education matters). He reasons with simple models held as useful, not true, and across uncertain trajectories. He borrows contested ideas as provocations and flags their limits. He turns threat into possibility (“learn how to be human”). He holds a plural idea of value, including awe and wonder, grounds steering over stopping in his view of human nature, and names the risk of blind devotion to the status quo.
2025-04-01 when-ai-takes-the-wheel. It shows little on its own, but it adds three things. The keynote becomes speculation in conversation (“we also get speculative”). He experiments with AI in public while keeping authorship (“I would not have chosen “brain mush!””). And the post completes the talk, post and podcast cycle through which his ideas develop.