B06 perspective notes: how he thinks (13 posts, 2021-03-28 to 2022-12-13)#

Read for method, values and role rather than for concepts. All 13 posts were read in full. There are no Modem Futura posts in this batch; 2021-05-28 announces the Future Rising podcast, a different show. Every quote below was checked against the corpus text (apostrophes and quotation marks normalised to straight form).

Provenance and what counts as evidence#

• Excluded as evidence: ChatGPT’s four answers in 2022-12-08 (only his questions, verdicts and framing count). The Midjourney images in 2022-09-16 (only his narration of the process counts). The “ChatGPT/Dall-E” header image on 2021-04-09, added later. Other people’s words: the Laurie Lee passage in 2021-04-09 and the Steve Tilley quote in 2022-12-13.
• 2021-10-05 is mostly an excerpt from Films from the Future, ch. 2 (Jurassic Park), which is his own prose written around 2017. It is wrapped in a new 2021 introduction and addendum. I do not have FFTF page numbers, so it is cited here as “FFTF ch. 2, as reposted”.
• Many posts are scholarship moved into public view:
• the first FRANKx lecture at ASU (2021-04-09);
• a National Academies / Los Alamos / NNSA symposium talk (2021-08-03);
• opening remarks to an ANSI Nanotechnology Standards Panel workshop, August 2020 (2022-02-10);
• a lecture video from his course FIS 338 (2021-06-09);
• drafts of articles for The Conversation (2021-09-07, 2022-06-13).
• Context: he is writing as Associate Dean in ASU’s College of Global Futures (2021-05-28). By September 2022 the Future of Being Human is “a new initiative” whose website supplies his Midjourney prompt (2022-09-16).

Posts in this batch (date, slug)#

1. How he thinks here#

He enters through a popular moment and then turns attention away from the obvious fear. Many posts start from a cultural event: a Musk product launch, a Super Bowl ad, a blockbuster film, a Nobel Prize, a new AI tool, a Radio Canada news story. He takes the event seriously and then moves attention from what is salient to what matters more. - Tesla Bot: The subtitle concedes the robot “looks like something out of a dystopian sci-fi movie. But that’s the least of our problems.” The post ends on “deciding who gets to imagine the future, and be a part of sharing in building it?” (2021-09-07 should-we-be-worried-about-elon-musks-tesla-bot). - Alexa ad: The joke about a mind-reading Alexa becomes a way into data fusion, predictive inference and “a seeming-quest to make us all ever-more-dependent super-consumers” (2022-02-12 scarlett-johanssons-amazon-alexa). - Dominion: A film he calls “not a good movie!” becomes a “great conversation opener” about the gap between could and should (2022-06-13 jurassic-park-dominion).

The move is never contrarian for its own sake. He first grants the obvious reading (“It was, perhaps, inevitable”) and then shows where the more consequential issue lies.

Stories and anecdotes do real work in the argument. - Laurie Lee: The FRANKx lecture opens with Lee’s village, where the horse’s “eight miles an hour” was “the size of our world, our prison.” At the close, that image becomes his question: have “the conventional ways we think about science, technology, and innovation” become our prison (2021-04-09 bounded-infinities)? - The piece of string: The standards remarks open with the rumoured Oxbridge exam question “how long is a piece of string?” Even this comes with a caveat: “to be honest l don’t know how true these rumors were” (2022-02-10 are-we-asking-the-right-standards-questions). - The snake: In the FFTF excerpt, a detail from the Arkema plant logbook carries the systemic point about cascading failure: “Overflowing toilets and snakes? Probably not.” (2021-10-05 butterflies-chaos-theory, FFTF ch. 2 as reposted).

He reasons by structural analogy, playfully, and checks where each analogy breaks. - Base coding: He sets digital, biological and material “base codes” side by side and treats “transcoding” between them as the real game-changer. He also warns “we need to be careful that we don’t fall into the trap of assuming DNA is the biological equivalent of digital code,” and says “straightforward analogies with digital base code don’t work.” - Quantum tunnelling: “I am using this as a metaphor, no more,” and later “at the risk of pushing the analogy to breaking point.” - Possibility spaces: He uses mathematical play (a universe of odd numbers that cannot reach the even ones; the Mandelbrot set as a plane of infinite but bounded possibilities) as a thought experiment about thinking itself (2021-04-09 bounded-infinities).

The physicist supplies the imaginative material, and the physicist’s discipline marks where the metaphor stops holding. The pairing of play with rigour about the limits of play is characteristic of him.

He asks first-principles and purpose questions. - Standards: “Why are we considering developing standards related to advanced materials in the first place?” and “What evidence is there that nanotechnology-related standards have led to measurable positive outcomes?” He also strips away labels to get to behaviour: “nature doesn’t care what we call a material, it just cares about how it behaves” (2022-02-10). - Graphene masks: The four “basic no-brainer questions” are: can it get into the body, can it cause harm, what harm and how, and how much is needed (2021-03-28 how-safe-are-graphene-based-face-masks).

He reasons carefully under uncertainty and tests plausibility. The graphene post is a worked example of how he reasons when the evidence is incomplete. He builds a chain step by step: 1. the analogous material (carbon nanotubes) and what makes it harmful; 2. how graphene differs (“But safer doesn’t mean safe.”); 3. the relevant dose metric (number or surface area, not mass); 4. deposition physics: single platelets are probably not dislodged, but clumps could be, and “any released airborne particles up to around 5–10 µm in diameter could potentially present a health hazard”; 5. the search for release data: “Here, I must confess I’ve hit a dead-end … But this in itself is a red flag.”

The finding is the absence of evidence where evidence should exist.

The same calibration appears elsewhere: - Alexa: He moves from existing features (Alexa Hunches, research on voice biomarkers) to what is “entirely plausible,” and flags “Of course, much of this is speculation.” (2022-02-12). - Chaos theory: He values it because its boundaries separate “plausible futures from sheer fantasy” (2021-10-05).

He holds tensions instead of resolving them early. - Graphene: He hopes the risks prove negligible (“I sincerely hope that any risks … will be negligible”) and still calls marketing without data “irresponsible innovation on a grand scale — even if the risks turn out to be negligible” (2021-03-28). - The solution problem: Innovation’s track record “isn’t necessarily a bad thing” on average, but “average” “hides a multitude of sins in the tails of the distribution” (2021-04-09). - Musk: His plan “makes sense” and “could open up a future full of promise for billions of people.” The concern that remains is whose future it is (2021-09-07). - Gene editing: The responsibility record is good, but “a community of well-meaning scientists and engineers are unlikely to be sufficient” (2022-06-13). - Chaos: Unpredictability is real, but so are “points of stability” and futures that “can be squandered” (2021-10-05).

He builds or experiments to find out. - Midjourney: Weeks of “playing around,” then a public account of the process (2022-09-16). - ChatGPT: Within a week of its release he ran a “quick and dirty experiment,” choosing to test it on his own field: “where better to start than with my own field” (2022-12-08). - AI risk literature: He did a bibliometric check on his own hunch and reported that his first reading was wrong: “I was initially pleasantly surprised by the data … on closer examination it became apparent” that most papers were about using AI to manage other risks (2021-08-03). - Dominion draft: He published the pre-editorial draft so readers could judge “how the process of working with an editorial team changes the tone and feel of a piece” (2022-06-13). - Podcast: He launched a podcast to find out whether his voice survives the move from page to ear (2021-05-28).

The experiments are small, quick, public and reflexive.

Curiosity, play and serendipity drive the thinking. - The lecture: He met the FRANKx brief “with some relish” and “took the brief quite seriously!”, organising it around “bounded infinities” and “metaphorical quantum tunneling.” Part Five is titled “The Art of Serendipity.” There he says what fascinates him is “how the juxtaposition of seemingly unrelated ideas can jolt us out of conventional ways of thinking.” Future Rising, a “mosaic of sixty short and often deceptively simple reflections,” was built to “stimulate the serendipitous emergence of new perspectives” (2021-04-09). - Midjourney: Serendipity by design. The prompt was “a sentence I somewhat randomly chose” from the initiative’s website (2022-09-16). - ChatGPT: He notes “something of a serendipitous irony” in AI helping to explain the concepts needed to govern AI, “But it’s one we should be running with.” (2022-12-08). - Remembered delight: As a physics undergraduate in the 1980s, fractals and strange attractors were everywhere, and kids would “revel in its complexity” (2021-10-05). - Film: Idle curiosity starts the film list: “I was curious to see how, if at all” the landscape of tear-jerking AI films had changed (2022-12-13).

In 2021-04-09, play is the argument, not decoration. Putting unrelated ideas side by side is the mechanism he proposes for escaping bounded, conventional thinking.

2. What matters to him#

People’s ability to shape futures they want, together. This is the commitment that recurs most often in the batch, and it goes beyond avoiding harm. - A future “more just, equitable, sustainable and vibrant for as many people as possible, rather than simply being different from the past”; learning to “build the future we aspire to rather than the one we’re otherwise destined for” (2021-04-09). - “how we build a more just, equitable and vibrant future together” and “how we help everyone to be a part of this” (2021-05-28). - The danger that “we risk being robbed of the futures we aspire to,” and that we hand our futures “to innovators whose vision exceeds their understanding” (2021-09-07). - “where science and technology benefit as many people as possible” (2022-06-13).

What he wants to protect is agency over the future, as much as health or safety.

The benefits are real, and he wants them. Used well, the Tesla Bot ideas are “transformative ideas and technologies.” Mastery of biological code gave us mRNA vaccines (2022-06-13). ChatGPT is “a potential game changer” for communication (2022-12-08). Midjourney can “unleash creativity rather than diminish it” (2022-09-16). He does not treat technology as the problem.

Responsibility means forethought, and it is judged by process as well as outcome. - “Fortunately, many products of nanotechnology are relatively safe — or can be rendered safe with some forethought.” - “But when it comes to the risks of using new technologies, hope alone is not good enough.” (2021-03-28).

Trust, and the ordinary people who do the right thing. In the graphene post, the harm he names includes eroding “trust and confidence in mask use” and “raising anxieties within those who have been using face masks responsibly” (2021-03-28).

Not forgetting what is already known. - He is angry at “a level of naivety and disdain for past risk research.” His irony: “hundreds of millions of dollars” spent on nano-risk research, yet “no-one seems to be asking the questions that count” (2021-03-28). - In the standards remarks he recovers a 1950s aerosol-science tradition that nano standards had drifted away from: “Interestingly, there was even discussion over half a century ago…” (2022-02-10).

Things that frustrate him: - labels treated as science; - opaque regulatory claims (“frustratingly difficult to find out what exactly FDA approval means”); - expert prose “conveyed in such convoluted ways by experts that they become impenetrable” (2022-12-08); - “the elitism that still shrouds higher education” (2021-04-09); - being “so enamored by the glitz of new tech” that we miss the bigger picture (2021-09-07); - a few powerful visionaries “impose their visions of the future on us” (2021-09-07).

Things that delight him: - physics and fractals; - being “blown away” by Midjourney; - ChatGPT’s clarity; - films that make him cry, including After Yang on a long flight, “to my embarrassment … tears running down my face!”; - a film he has “a soft spot for” despite the critics (Bicentennial Man); - a lecture brief he could take on “with some relish.”

Humour is part of his register. Examples: “an army of racist, sexist Tesla Bots”; “if Hammond had read his Perrow”; images made “for a poster to go above the office coffee machine!”; “mindless entertainment :)”.

He lets feeling show. “I must confess that I’m concerned by what I see unfolding” (2021-03-28). “I find these both inspiring and thought provoking” (2022-09-16).

3. Risk as a way of thinking#

Novel technologies need a new mindset. His strongest statement of this in the batch is the FRANKx lecture (2021-04-09). He does not put it in risk vocabulary, but it is the reasoning underneath his later risk vocabulary. - The old domain: Historically, the consequences of innovation arrived more slowly than our ability to innovate fixes. In that domain “trial and error have consistently been shown to be an effective problem-solving technique, as there’s always time to learn from ‘trials’ that don’t quite work as expected.” - The new domain: When consequences outpace fixes, “simple models of innovation do not work within this domain.” We will have to depart “quite radically in some cases” from conventional thinking. His conclusion: to navigate what this creates, “we’ve got to change our mindsets, change our skillsets, and change the ways that we think about the future.” - Why: The driver is our growing ability to “base code” across digital, biological and material domains, with intelligent machines produced “through a merging of all three.” - What the image does: “Bounded infinities” pictures conventional, stovepiped thinking: infinite options inside a frame that leaves out the ones we need.

In effect, this is an argument for why predict-assess-fix risk management runs out of road, and why the change has to be in mindset rather than method.

Risk Innovation at work: a way of seeing that starts from value (2021-09-07 Tesla Bot). This post uses the full vocabulary: - “risk landscape”; - “orphan risks — risks that are hard to quantify and easy to ignore, and yet inevitably end up tripping innovators up if overlooked”; - asking “how it might threaten what’s important to them, and how to navigate these threats”; - the three questions: “What’s important to your enterprise, your investors, your customers, and the communities you touch? What orphan risks threaten these things of importance? And what small steps can you take to help navigate this complex risk landscape?”

Three points about how these work: 1. They start from value, not hazard. That is why privacy, autonomy, identity, how people respond to humanoid robots, and ideological mismatch all come into view. 2. They are for everyone. They are “questions we should all be asking as powerful entrepreneurs strive to impose their visions of the future on us,” not a procedure for specialists. 3. “Small steps” is humble. It means steering, not a master plan.

He puts it plainly: “The hard truth is that developing new technologies in socially responsible and beneficial ways means navigating an increasingly complex risk landscape where social and political awareness are paramount, and technological competency is often the least of a developer’s problems.”

Nuance: the vocabulary is still shifting. In the same post he calls the orphan risks “manageable” (“But complex as these threats are, they are manageable.”). What cannot be managed is “the deeper concerns”: power, and who gets to imagine the future. In 2021 the language of navigation sits on top of management terms rather than having replaced them.

Early positioning on AI (2021-08-03). The title pairs “navigate” with “risks and benefits.” The slide caption reads “navigate complex threats and grow/protect value.” The subtitle asks whether we are “so obsessed with the rights and wrongs of AI that we’re taking our eye off potential risks?” - His evidence: ethics guidance had grown fast, while work on “the risk-implications of artificial intelligence” was thin once papers using AI to manage other risks were filtered out. - What he means by risk here is not narrow technical assessment. It is “innovative approaches to risk” that help “ensure that AI enhances the future opportunities we’re facing rather than diminishes them.” Risk is tied to what could be lost and what could be gained. - The text is short; the substance is in the talk video, so this is a position statement rather than a full argument.

ChatGPT, week one (2022-12-08). “Navigating this risk landscape is going to need new thinking,” including how to “ensure the growth of social as well as economic value.” The risks he names are largely orphan-type risks, not conventional hazards: “the dangers of unhealthy attachments and gullibility,” bias, misinformation, disinformation and hate speech. These are “a whole raft of risks that we’re just beginning to unpack.” This is humility about a poorly understood technology, not a verdict on it.

The empirical roots of the new-mindset idea: nanotoxicology (2022-02-10). His own research experience is that novel behaviour escapes conventional risk categories. - Conventional metrics (bulk chemistry, mass) missed how particle size, surface area and surface chemistry drive harm. - The 2011 “sophisticated materials” categories pick out materials “likely to slip under the conventional risk radar.” One category is materials whose hazards are “not adequately captured by conventional hazard assessments.”

He is equally firm that novelty alone is not the test: - “there is no reason to assume that new materials, by default, present new risks”; - “existing materials that are used in new ways can lead to unexpected risks just as readily.”

The test is behaviour and function, not what a thing is called.

He renames things on purpose, and knows he is doing it. He coined “sophisticated materials” “to try and break away from the conceptual baggage that comes with the terms ‘nanomaterials’ and ‘advanced materials.’” He adds, with a smile, that it is “yet another term of art” and “didn’t do us any favors citation-wise!” This is the same move his later risk vocabulary makes: new terms as mental tools for seeing differently, held with awareness that they are terms of art, not science.

Humility against false precision (2022-02-10). “100 nm, no matter how useful it is, is a number of convenience, not of science.” Treating a term of art as a term of science is “akin to asking how long is a piece of string and trying to convince people that you have a definitive answer!” and risks “a house of cards.”

His quantitative foundation is still in use (2021-03-28, the same fortnight as the FRANKx lecture). The graphene post draws on exposure science, dose metrics, alveolar deposition and particle-size limits, plus “the universal rule of ‘better safe than sorry’” when mechanisms are unknown. The harms he counts go beyond toxicity to lost trust and anxiety, so a threat to value is already part of an exposure-science post. The foundation is built on, not discarded.

Complexity is the logic of navigation (2021-10-05, FFTF ch. 2). He accepts that “we cannot wield perfect control over complex technologies within a complex world,” and Perrow’s normal accidents make surprise inevitable. But complex systems also have bounds and “points of stability.” Some futures “are more likely to occur if we take the appropriate courses of action,” and others “can be squandered.” That is the intellectual basis for navigating rather than predicting or controlling. He echoes it in 2022-06-13, where downsides are “anticipated and navigated.”

Summary. In this batch the risk concepts mostly reorient attention. They ask: - what matters here; - what is being overlooked; - who decides; - which futures are being opened or closed.

They work as procedures only secondarily. Where conventional tools work, as with aerosol exposure, he uses them precisely.

4. Scholarship and public writing#

The posts are his scholarship in public. Lectures, symposium talks, standards-workshop remarks, a book excerpt, class videos and drafts for The Conversation are all reposted here. The standards post cites his Nature (2011), J. Nanoparticle Research (2005) and Toxicological Sciences (2011) papers in footnotes (2022-02-10). The blog is where a peer-reviewed argument is voiced again for a new audience and a new problem (advanced materials).

He shows work in progress, openly. - The lecture notes are “still very much a work in progress and so are a little rough around the edges.” Mid-lecture he admits he has “laid out a vision of the challenges we’re facing with very little supporting evidence” (2021-04-09). - The graphene post is labelled “This is an emerging story” and carries dated updates. The August 2021 note that Health Canada had re-permitted four Shandong models is placed near the top rather than buried (2021-03-28). - The film list was “Updated September 2023” (2022-12-13).

How he handles evidence and expertise. - He uses his credentials sparingly, and to justify concern, not to close debate: “I’ve been at the forefront of researching nanomaterial risks … for over 20 years” (2021-03-28). - He marks the edge of his competence: “it lies outside my current work — I’d hope though that participants in this workshop can” (2022-02-10). - He credits others’ work, including a civil-society critic: Jim Thomas of the ETC Group “for bringing this issue to my attention and doing some of the initial digging” (2021-03-28). - He is frank about his own field’s failings: “believe me, this is my field, I know!” (2022-12-08).

Transdisciplinarity is his normal way of working, not a gesture. Within one batch: - aerosol physics and toxicology; - standards policy; - quantum mechanics and fractal mathematics as metaphor; - Laurie Lee and Ray Bradbury; - Charles Perrow; - Buddhist koans (“to jolt us out of the ruts of conventional thinking”); - “the arts and humanities” and “alternate ways of knowing”; - science-fiction film; - AI image-making.

His educational ideal holds both sides: “maintaining much of the deep knowledge associated with traditional disciplines” while “transcending these disciplines” (2021-04-09).

Accessibility and form are experiments in reach. - He calls himself “a very much an ‘aural’ writer,” worries “how much gets lost in translation” between how he hears his prose and how it reads, and launches 3–5-minute podcast episodes to “explore new ways of opening up my work to a broader audience” (2021-05-28). - He posts his class video because it “distills a lot of research and thinking” into something he hopes is accessible to anyone (2021-06-09). - He praises ChatGPT because clarity for “a broad audience” matters to him (2022-12-08). - He is interested in how editing changes a piece’s “tone and feel,” which shows how much voice matters to him (2022-06-13).

Teaching belongs to the same project. The Moviegoer’s Guide to the Future (FIS 338) “conveniently uses science fiction movies as a starting point” (2021-06-09). Jurassic Park is “a staple” of his undergraduate classes (2022-06-13).

5. His role as he sees it#

A catalyst, not an oracle. This is his clearest statement of purpose in the batch. He is fascinated by “how the juxtaposition of seemingly unrelated ideas can jolt us out of conventional ways of thinking … This is exactly what I set out to achieve in much of my writing.” Future Rising “was specifically conceived and designed to catalyze metaphorical quantum tunneling in its readers.” He closes the lecture “in the spirit of taking on the role of quantum tunneling catalyst” by leaving three questions rather than conclusions (2021-04-09). His writing is meant to move readers into new ways of seeing.

He asks questions. Many posts end on a question: - “how do we know we’re asking the right questions with advanced materials?” (2022-02-10); - “is this enough to keep Amazon from going there?” (2022-02-12); - who gets to imagine the future (2021-09-07); - the three education questions (2021-04-09).

He treats readers as co-thinkers and leaves the last step to them.

An expert who raises a flag when the evidence is missing, carefully. In the graphene post he names “irresponsible innovation” while insisting “we don’t yet know if graphene particles are being released” (2021-03-28).

He works across sectors and audiences: - National Academies, Los Alamos and NNSA (2021-08-03); - the ANSI standards community (2022-02-10); - entrepreneurs and investors through the Risk Innovation Nexus (2021-09-07); - undergraduates (2021-06-09, 2022-06-13); - readers of The Conversation; - civil-society researchers (2021-03-28).

He is fair to the people he criticises. - He does not tell Musk to stop. He writes: “this is not to say that Tesla Bot isn’t a good idea, or that Elon Musk shouldn’t be able to flex his future-building muscles in this way” (2021-09-07). His criticism is aimed at power over the future, not at the person. - The gene-editing community gets credit (“Fortunately, consideration of the social and ethical side of gene editing has gone hand in hand with the science’s development”) (2022-06-13). - Nano standards work is not dismissed: “Of course, I’m oversimplifying … I don’t want to downplay the importance of work that has been done” (2022-02-10). - Speculation about Amazon is put as a question (“Is Jeff Bezos hiding his long-term plans in plain sight…?”), not an accusation (2022-02-12).

What he refuses to do. - Fear-mongering: Killer-robot tropes are “the least of our problems” (2021-09-07). A whole post pushes back against dystopian AI films with ones that “hint at a future where smart machines can and do bring out the best in us” (2022-12-13). - Hype: “Beyond the hype and the fear” (2022-12-08). - False precision: “a number of convenience, not of science” (2022-02-10). - Preaching: He ends on questions. - Polarisation: He holds both sides throughout.

He is not neutral, though. He takes clear stands on irresponsible innovation, on concentrated power over the future, and on elitism in education.

His relationship with readers is confessional, warm and funny. “I must confess” is a recurring phrase: - on graphene: “I must confess I’ve hit a dead-end”; - on standards: the “sophisticated materials” name; - on Midjourney: being “blown away”; - on the podcast: being an “aural” writer; - on the films: “a somewhat personal list.”

He admits tears, embarrassment and soft spots. He also invites readers in: “how do they compare?” (2022-06-13), “Happy listening!” (2021-05-28).

He corrects himself and holds his views provisionally. - He revised his first reading of the AI-risk literature (2021-08-03). - The chaos addendum: “Looking back at my work, there’s always stuff I wish I’d included or written differently,” with regret at leaving out Bradbury’s A Sound of Thunder, ending “nothing is certain in the complex world we live in!” (2021-10-05). - The graphene update records the regulator’s reversal (2021-03-28). - On AI art: “Based on my experiences so far, I’m not sure I’m worried” (2022-09-16). - ChatGPT was “a quick and dirty experiment,” and he is explicit about how early his judgement is (2022-12-08).

He wants futures-building opened beyond the academy. He wants to tunnel “beyond the walls of formal higher education” and “empower anyone who’s interested to be part of a positive future-building enterprise” (2021-04-09).

6. What is distinctive#

1. Rigorous risk science and deliberate play in one mind, in one fortnight, each keeping the other honest. On 2021-03-28 he reasons about respirable graphene clumps “up to around 5–10 µm.” On 2021-04-09 he reasons with odd-number universes, fractal planes, quantum tunnelling and koans. Few people writing about AI can do both. Fewer still use the physicist’s precision to police the physicist’s metaphors (“I am using this as a metaphor, no more”).

2. Missing evidence as a finding, and responsibility judged by process. He calls marketing without data “irresponsible innovation on a grand scale — even if the risks turn out to be negligible.” When evidence that should exist does not, that is itself the red flag. This is not how most technology commentary works, since it usually waits for demonstrated harm or dismisses the concern.

3. Labels versus behaviour. “Terms of art” versus “terms of science,” “nature doesn’t care what we call a material,” and “no reason to assume that new materials, by default, present new risks” (2022-02-10). My inference, not his claim in this batch: the move carries straight over to AI governance, where “AI,” “frontier” and “general-purpose” are terms of art. It points towards governing by what systems do and where they reach, not by definitional category, and without assuming either that new means dangerous or that familiar means safe (compare “it’s tempting to think of carbon as being safe, just because we’re familiar with it,” 2021-03-28).

4. A third position on AI in 2021. He found AI discourse rich in ethics guidance and thin on risk. What he proposes is neither existential-risk safety nor compliance. It is navigating a risk landscape anchored in value and future opportunity (“grow/protect value”), starting from what matters to the people and communities involved (2021-08-03, 2021-09-07).

5. A structural reason why novel technologies need a new mindset. The innovation–consequence timescale inversion: once consequences outpace fixes, trial and error, “simple models of innovation” and “sciencing” our way out stop working. AI appears as one product of transcoding across base codes, not as a separate topic (2021-04-09). This is a systems argument, not hype or alarm.

6. Science fiction used in three ways at once. Films work as serious thinking tools (the Jurassic Park chapter; FIS 338). Their surface fears are moved onto questions of power and agency (“who gets to imagine the future”). And they are a source of empathy and hope, where AI “holds a mirror up to our own humanity” (2022-12-13). Most commentators use film either as illustration or as a scare.

7. Early, hands-on, public, delighted experimentation with generative AI, with risks named in the same breath. - He frames making images as “co-creation,” “with me prompting Midjourney and in turn being prompted and inspired by what was returned,” while keeping meaning with humans: “the interpretation is all on my side here, not the AI’s” (2022-09-16). - A week after ChatGPT’s release he flagged “unhealthy attachments and gullibility” as risks (2022-12-08). That was early for these to be named as a primary concern.

8. Expertise with a conscience about communication. He is willing to say that a machine explained his own field more clearly than most experts (“believe me, this is my field, I know!”). He sees clarity and access as part of responsible innovation, not a public-relations extra (2022-12-08).

7. Posts that best reveal how he thinks#

1. 2021-04-09 bounded-infinities-quantum-tunneling-and-the-future-of-education. His own statement of method and purpose. It covers: - the solution problem and the timescale inversion (why novel technology needs a new mindset); - structural analogy with its limits flagged; - play and serendipity as the way out of conventional thinking; - writing designed to catalyse readers; - education beyond the academy.
2. 2021-09-07 should-we-be-worried-about-elon-musks-tesla-bot. Risk Innovation as a portable, value-first way of seeing: risk landscape, orphan risks, three questions, “small steps.” It also shows him setting aside the sci-fi fear to reach “Whose Future?” (his section heading), and being fair to Musk.
3. 2021-03-28 how-safe-are-graphene-based-face-masks. His quantitative foundation at work. Step-by-step reasoning under uncertainty, missing evidence as the red flag, responsibility judged by process, harm to trust counted as harm, and live public updating and correction.
4. 2022-02-10 are-we-asking-the-right-standards-questions-about-advanced-materials. First-principles questions about purpose, terms of art versus terms of science, humility against false precision, behaviour over labels, novelty not assumed to mean risk, and self-aware renaming to shed conceptual baggage.
5. 2022-09-16 56-stunning-ai-generated-images-inspired-by-the-future-of-being-human. Play, curiosity and serendipity in practice. He experiments in public, shows the process as well as the product, frames the work as co-creation, and holds his conclusions provisionally.
6. 2022-12-08 i-asked-open-ais-chatgpt-about-responsible-innovation. A fast experiment in his own field. It shows how much he cares about accessible explanation, uses navigation and value language about AI, and names early relational risks (attachment, gullibility) alongside enthusiasm.

Also worth a look: the 2021-10-05 addendum, a small, candid self-correction that ends on uncertainty, and 2021-08-03, his early diagnosis that AI discussion was rich in ethics and poor in risk thinking.