B03 perspective notes: how he thinks (14 posts, 2018-11-25 to 2019-06-12)#

Batch context. These are the months after Films from the Future (FFTF, Mango, November 2018) came out, while he was directing the ASU Risk Innovation Lab and teaching in the School for the Future of Innovation in Society. Much of the batch is republished or reworked: pieces from The Conversation (one from December 2015), OneZero and LinkedIn, and extracts from FFTF, which he wrote alone. All of it is his own prose. There are no AI-written, guest or co-written sections and no Modem Futura posts. Two posts are light pieces (a satirical presentation guide and an Oscars list). I kept them in because they show his play and his values about communication more directly than the serious pieces do.

Post key (cited below by date and slug stem):

1. How he thinks here#

He opens with an oddity, a concrete case or a place, not with a thesis. Almost every post starts from something specific that doesn’t quite add up, and the puzzle becomes the way in. - A technology booster gives Elon Musk a “Luddite Award” on the day SpaceX lands a rocket: “It’s an odd juxtaposition, to say the least.” … “So what’s the deal?” (2018-12-15 if-elon-musk-is-a-luddite). - Google’s AI ethics council collapses within nine days: “Ironically, Google’s new group of ethics advisors fell apart because of ethical challenges” (2019-04-15 tech-companies-need-an-ethics-reset). - A headline about self-driving cars and skin tone: “I was struck afresh by how similar the challenges … are” to chemical risk (2019-03-05 should-we-be-treating-algorithms). - Fitbit launches a tracker you can’t buy in a store (2019-03-01 personal-fitness-tracking). - His wife is “aghast” at the word “morality” in his book title (2019-04-26 avengers-endgame). - He watches young researchers give “earnest but painful” talks (2019-02-08 how-to-give-the-best-scientific-presentation). - The closing chapter of FFTF begins with him on the Isle of Arran, “looking out over the Firth of Clyde” and grappling with “patchy Wi-Fi” (2019-03-06 navigating-the-risks-and-benefits; FFTF p. 287).

The general point comes out of a particular that surprised, irritated or charmed him.

Reframing is his main move, and he often does it by finding a gap in an existing frame. - Orphan risks start from Rumsfeld’s three kinds of uncertainty. He notices what that taxonomy leaves out: risks that are “‘known knowns’ if you’re looking in the right place” but get neglected (2018-12-13 tech-startups-orphan-risks). The new idea sits in the blind spot of an old one. - Luddite is reframed as “Being cautious ≠ smashing the technology” (2018-12-15). The binary he inherits (pro- or anti-innovation) is refused rather than answered. - Ethics is reframed from principles to outcomes, “the often tortuous pathway between aspirational goals and what happens when the rubber hits the road” (2019-04-15). - Designing the future is reframed as a question of who gets to do it: “the future is designed by the powerful” (2019-03-31 design-principles-for-de-marginalizing). - Sci-fi is reframed from prediction to imagination. Films help “not because they are accurate or prescient, but precisely because they are not tethered to scientific accuracy” (2019-03-06; FFTF p. 288). He also names “dare I say it, their entertainment value” as part of how they work.

He argues by structural analogy, and marks where the analogy stops. The algorithms-and-chemicals post is the clearest example of his method in the batch (2019-03-05). - It opens by conceding the difference: “At first blush, algorithms and hazardous chemicals have precious little in common.” Then it looks for shared structure: both “impact our lives in ways that are neither desirable, nor always easy to predict.” - He states the limit plainly (“Of course, an algorithm is not a chemical”) and then argues that once you look past the different mechanisms, “the analogy between algorithms and chemicals becomes intriguingly compelling.” - He then builds new concepts by translation, thinking aloud as he goes. Is there an algorithmic equivalent of exposure? “I think there is.” That yields “algorithmic exposure” and an “algorithmic exposure-response relationship.” - The analogy is used to produce thinking, not to prove anything. He ends by limiting it again: chemical methods “are not nearly as robust as we’d like them to be.”

He tests plausibility against the physical world. In the same post he asks whether the lab finding would actually carry over to on-road cars, which use LIDAR and whole-object detection, and concludes there is “no clear algorithmic exposure route”. He adds straight away: “This in no way negates the research.” His FFTF material makes the same instinct a general heuristic: Occam’s razor for “grandiose claims and fearful projections” (2018-11-25 twenty-trends). He calls himself “something of an agnostic” on superintelligence and observes that its promoters behaved as if at “a religious convention” (FFTF p. 170, excerpted in 2018-11-25).

He holds tensions and doesn’t resolve them early. Many posts are built around a “Yet”. - Nostalgia and obligation. Arran’s slower life comforts him, yet that comfort is a “sentimental illusion”, and renouncing technology “from a position of privilege” denies others choices (2019-03-06; FFTF p. 288). - Acceptable and zero risk. The ideal is “no such thing as acceptable risk”, but society has had to accept there is “no such thing as zero risk”. On skin-tone bias there is “a very strong case” that no risk is acceptable, yet human drivers show a similar bias. So even here there is a trade-off with the status quo, and “there’s always a risk of not developing or using a product” (2019-03-05). - Ethics necessary and insufficient. Don’t “throw the ethics baby out with the bathwater”, but ethics alone “are worth little without mechanisms” (2019-04-15). - Dual use. A lip-reading AI that could help people who have trouble speaking could also enable mass surveillance (2019-01-16; 2019-03-07). - Both sides of Musk. He defends Musk against the Luddite charge but faults OpenAI’s founding logic, “the answer to technology innovation is… more technology innovation” (2018-12-15). - “Don’t Panic”, but not complacent (2019-03-06).

Curiosity and delight drive the choice of material. - He is openly intrigued: “I’ve become increasingly intrigued” by what films reveal. “Intriguingly, part of the answer here could lie in … watching more science fiction movies” (2019-01-16). - The “twenty trends” list roams from de-extinction to extraterrestrial life and includes Occam’s razor as a “trend”. The book extracts show a scientist taken with audacious ideas: Roger Angel’s space sunshade “caught the imagination of a number of us who were intrigued by such an audacious approach” (FFTF p. 265). - He runs small experiments to find things out. “I recently conducted an informal poll in my undergraduate class”: over half admitted using smart drugs (2018-11-25).

He says, in effect, what his creative method is. Explaining why The Hitchhiker’s Guide frames the end of FFTF, he admires “the skill with which Adams creatively melded together odds and ends of ideas from very different places to create new ones” (2019-03-06; FFTF p. 289). This is also a good description of his own practice in this batch: - Rumsfeld and startup finance - toxicology and machine learning - the Beatitudes and design guides - Oppenheimer and the Marvel universe - Douglas Adams and a book on emerging-technology governance

Recombination across distant domains is the method, not decoration.

Play is part of how he argues. - The presentation post is a satire, an “anti-guide”, written “in my frustration” (“Jargon is your best friend”). By “popular demand” it grew a PowerPoint Bingo Card (2019-02-08). - The Oscars post interprets science “rather liberally”, coins “tech-fi” as a genre he thinks deserves recognition, and cannot resist a joke about A Quiet Place being nominated for Sound Editing: “and no, the irony is not lost on me!” (2019-02-24). - FFTF “fails miserably on the ‘where to get the best drinks’ front” (2019-03-06). - The film-list captions are posed as questions to the reader: “If a simple pill would boost your smarts, would you take it?” “When you are part-machine, who owns you?” (2019-06-12).

The playfulness does serious work. It lowers the barrier to entry, and the satire carries a real critique of how science is communicated.

2. What matters to him#

Who gets to shape the future. The strongest value in the batch is distributive and democratic. - Innovation must improve “the lives and livelihoods of all — not just the privileged” (2018-12-15). - The poor, the homeless and the “educationally challenged” have as much claim to the future as the powerful. Without inclusivity the default is “a future of the powerful, by the powerful, for the powerful” (2019-03-31). - Leaving these questions to “experts” is “an abdication of responsibility”. We should steer toward “the future we want, rather than one that someone else decides for us” (2019-03-06; FFTF p. 288). - Elysium is recommended for how “AI and robots impact rich and poor communities differently” (2019-06-12).

People’s dignity and control over their own lives. - His list of things risk can threaten includes “autonomy, dignity, trustworthiness, self-esteem, way of life, and even deeply held beliefs and convictions” (2018-12-13). - Algorithmic harms include threats to “livelihood … Or their liberty, dignity, and self-respect?” (2019-03-05). - The Fitbit worry is “handing over control of your life to someone else”, and he is blunt: “If this sounds creepy, it is” (2019-03-01).

An obligation to innovate, as well as caution. He is not a technology sceptic. “We have an obligation to explore new ways of using science and technology to improve the world,” an obligation that “comes with some tremendous responsibilities” (2019-03-06; FFTF p. 288). Technology can’t buy love and happiness (“you can’t simply ‘science’ your way to them either”), but it can make them “that much easier to achieve.”

Humility, as against the hubris of the brilliant founder. - The recurring villain is the “wealthy and unaccountable entrepreneur who is supremely confident in his own abilities” (Hammond, Nathan). The lesson he draws: “without humility and a good dose of humanity, our innovations can come back to bite us” (2019-01-16; 2019-03-07). - He applies the same humility to the species: “I’m not yet convinced homo sapiens have the tools and abilities to truly design the future”. Science reveals “a fractal-like depth to the complexity” in which certainty “all too often turns out to be illusion” (2019-03-31). - The FFTF chaos passage makes the anti-false-precision point historically. Lorenz overturned the assumption that better prediction came from “better data and better models” (FFTF p. 40, excerpted in 2018-11-25).

Honest use of evidence. He cares that evidence isn’t cherry-picked, whichever way the bias runs. - He respects hard-won method: the Cochrane Handbook is “long, it’s meticulous, it’s demanding. And it works.” Protocols that seem “onerous and pedantic to the uninitiated … save lives.” - He is generous about human weakness: “These are natural tendencies, and we’re all subject to them” (2019-03-05).

What frustrates him: - ethics-washing (“smoke-and-mirrors”, 2019-04-15) - good intentions that stay “naive and ephemeral” - scientists who have been “taught the opposite — how to communicate badly” (2019-02-08) - tech as a panacea - the sense that stories of hopeless dystopia leave us with the feeling “there’s not a lot we can do about it” (2019-03-06)

What delights him: - films, with real pleasure in the craft (Into the Spiderverse, “an awesome blend of science, technology, fantasy, and innovative story telling”) - Contact, which “beautifully captures the passion and thrill of what it’s like to be a scientist” (2019-06-12) - students’ changes of perspective (“I’ve been amazed”, 2019-04-26) - the “warm glow of vindication” when morality at the movies turns out to be “in” (2019-04-26)

3. Risk as a way of thinking#

This batch contains the earliest full public statement in the corpus of the cluster risk innovation / risk landscape / threat to value / orphan risk, in 2018-12-13. Read closely, it supports his September 2026 account in some places and complicates it in others.

Risk innovation is framed as a change of mindset. “In order to thrive in a world driven by innovation and the changes it brings, we need parallel innovation in how we think about and act on risk” (2018-12-13; he calls it a “deceptively simple idea”). The trigger is explicitly novelty: - New technologies are “rewriting the rule book on what is possible”. - There is a “growing chasm between tried and tested ways of thinking about risk” and the risks that now trip companies up. - Unprecedented capabilities “cannot be brushed off by assuming business as normal” (2018-12-15). - Companies need to “escape the ruts of conventional thinking” (2019-01-16) and “break from business as usual” (2019-03-07).

Threat to value is presented as something that opens a conversation, not as a formula. - He concedes it is “a somewhat subjective way of thinking about risk”. It “has the advantage of opening up conversations around factors that are often highly relevant to the decisions people make, and yet because they are hard to quantify, are frequently ignored” (2018-12-13). This is the clearest early evidence that he values these concepts because they make things visible and discussable. That is their use as mental models. - It “extends conventional thinking rather than replacing it”: health, wealth and environment “fit comfortably into the ‘value’ bucket.” - It has a reciprocal edge, “the reciprocal dangers of threatening what is important to others through what they do”. This reappears in the ethics post: “every action they took potentially threatened something of importance to someone else” (2019-04-15). - FFTF’s broad account of resilience fits the same outlook: it “gets to the heart of how we think about what’s important to us, and how we make provisions to protect and grow this” (FFTF p. 261). He speaks of protecting and growing value, not only defending against loss.

Risk landscape and navigation: risk as the terrain between intention and realized value. - The landscape is “the risk landscape that lies between new ideas and their successful implementation” (2018-12-13). - For Fitbit it is “an increasingly complex risk landscape that sits between the potential benefits … and whether benefits will be realized” (2019-03-01). - In Endgame it is “the tortuous landscape between good intentions and socially acceptable outcomes” (2019-04-26).

“Navigate” is the verb throughout, including the FFTF closing chapter’s title. Risk is the ground you cross to reach value, not a list of hazards to be controlled. Benefits are always on the map (“there’s always a risk of not developing”, 2019-03-05).

Orphan risks are a category the conventional taxonomy leaves out. It is developed in public over several months: - defined for startups and funders (2018-12-13) - “what we are increasingly calling ‘orphan risks’” (2019-01-16) - applied to consumer technology with a new twist: consumer trust can itself become an orphan risk of backlash, since “These kinds of risks aren’t a one-way street” (2019-03-01)

Where the 2018–19 record complicates the 2026 self-account. These concepts don’t yet work only as mind-opening models. 1. They are also presented as tools. The Risk Innovation Accelerator is “building tools” for startups and investors (2018-12-13). The ethics post calls for “measurable expectations, enforceable checks and balances” (2019-04-15). 2. The case is largely made in business terms (being “blindsided”, “the difference between success and failure”), with ethics “quite apart from” that. This reads as a deliberate choice to meet entrepreneurs in their own language rather than preach at them. 3. The most rigorous AI post in the batch goes the other way from “new mindset”. It argues that the young field of algorithmic risk should import the hard-won machinery of chemical risk assessment: - hazard versus risk - exposure - consequence - dose-response - weight of evidence

The field needs it “lest we end up building our understanding of algorithmic risks on an evidentiary stack of cards”, and “Unless, that is, we’re content to repeat the mistakes of the past as we reinvent the future” (2019-03-05).

That is the “foundations built on, not discarded” part of his account, visible early and in detail. Two features stand out: - He moves in both directions. Where old tools fail to see social risks, he asks for new thinking. Where a new field is naive, he asks for old rigour. - Even inside the quantitative frame he stretches the categories with his value lens: - “exposure” becomes anyone “impacted by the deployment of an algorithm” - “consequences” come to include “liberty, dignity, and self-respect”

So risk science and risk-as-value are already fused in one argument.

Complexity underlies the humility. Emerging technology, “like any complex system, … is likely to look great… right up to the moment it fails” (2018-12-15). The Day After Tomorrow caption makes the same point: “everything looks rosy … until it isn’t” (2019-06-12).

4. Scholarship and public writing#

The same ideas move between registers without seams. - Orphan risks first appears as a LinkedIn/Medium piece. - Risk innovation links to his academic paper (rdcu.be). - “Creative insights into the social landscape” links to his Nature Nanotechnology piece (doi 10.1038/nnano.2015.196). - The Conversation op-eds are reworked into posts. - The closing chapter of an academic-trade book is published as a blog post.

He revises one argument for different audiences. The 2019-01-16 and 2019-03-07 posts are near-twins, and the March version adds “or simply establishing an ethics board”, which anticipates the April piece. Ideas build up across posts and are refined in public.

Teaching is a place where he experiments and learns. - The entrepreneurial ethics course (Michigan, from 2013) was reshaped by what he found: “it quickly became obvious that these graduate students needed more than an ethical compass” (2019-04-15). His view that ethics must be operationalized comes from watching students’ “good intentions but no idea how to translate them into good practice.” - The ASU sci-fi course (“Each week, we watch and discuss a movie together”) is where he sees films change perspectives (2019-04-26). - The smart-drug poll is teaching turned into data (2018-11-25).

How he treats evidence and expertise. - He reads the primary source against its coverage. The predictive-inequity paper “is a little more guarded” than the MIT Technology Review headline or Kate Crawford’s tweet. He praises AI Now’s “sterling work” in the same post. - He is sceptical of single studies while keeping “a deep respect for individual studies that push the boundaries”. He is “suspicious (but not off-the-cuff dismissive)” of vested interests (2019-03-05). - He values expert communities: the responsible-innovation “thriving interdisciplinary community”, and his own WEF Top Ten Emerging Technologies role. But he refuses to let expertise close the question (2019-03-06). - Mitchell Baker is “right in part”. The humanities alone are not enough, and innovators need a “new formulation” of skills from many disciplines (2019-01-16).

He draws across many disciplines. Physics and materials science (nanotech as a continuation of materials science), toxicology and exposure science, complexity and chaos theory, ethics pedagogy, theology (a secular reworking of the Beatitudes, 2019-03-31), film criticism, Douglas Adams and comic books. None of these is treated as more serious than the others.

Accessibility is a value, not only a style. - The everyday analogies make technical ideas concrete: - bleach under the sink - gasoline in the car - a grizzly seen from a distance versus one met face to face - indigestion versus heart failure (2019-03-05) - The film posts give “What to watch for” and “Where to watch it” (2019-06-12). - His explicit rule: “no one ever reads an overlong, overweight, and utterly incomprehensible guide” (2019-03-06; FFTF p. 290). - The presentation satire is the same value put negatively: “the less understandable people find you, the smarter they will think you are” (2019-02-08). - Films are “social and educational levelers” that “build bridges between people who know how science and technology work, and those who know what’s needed to ensure they work for the good of society” (2019-01-16).

5. His role as he sees it#

A guide, not an oracle. The Hitchhiker’s Guide comparison is his self-description. It is a guide that “doesn’t even pretend” to explain everything and focuses on what travellers “really need to know”. FFTF is “a very incomplete guide”, and its aim is “how we think about technological innovation, society, and the future” rather than technical explanation (2019-03-06; FFTF pp. 290–291). Films help “as long as they’re accompanied by a trustworthy guide”. He notes that there are “remarkably few places for consumers to get trustworthy and relevant information” (2019-03-01).

Readers are fellow navigators, not an audience to be alarmed or instructed. - “We all need to be able to wrap our collective heads around what’s coming” (2019-03-06). - He invites rather than tells: “you’ll need to watch the movie for yourself … You may be surprised” (2019-04-26); “invite some friends round, open the popcorn and put on a movie” (2019-03-07). - Our real resources are “our collective intelligence, our ability to learn, our sense of right and wrong, and our capacity to work together” (2019-04-26).

With industry he is a critical ally. - He assumes good faith: “most entrepreneurs really are striving to make the world a better place” (2019-04-15). FFTF is “written with innovators in mind, who want to do the right thing and just don’t know where to start” (2019-03-07). - He is still sharp about ethics boards used “to blunt criticism” and about “equating ethics with success while skipping over the hard stuff in between.” - He speaks to startups and funders as an adviser, but his framing of success always includes affected communities and “the reciprocal dangers” to others.

What he refuses to do: - Fear-mongering and fatalism. “Don’t Panic” is “as good a piece of advice as any”. He warns that sci-fi without “scientific facts and social realism” breeds dystopian fatalism (2019-03-06). “Knee-jerk reactions to seemingly-startling results rarely result in socially beneficial outcomes” (2019-03-05). He flags Inferno for “the dangers of slipping into deeply unethical behavior while trying to address perceived global risks” (2019-06-12). - Hype and techno-salvation. He is agnostic on superintelligence, and Endgame “naively promotes the idea that, if handled right, powerful tech can be the solution to all our problems” (2019-04-26). - Polarization. He won’t accept the Luddite/innovator binary. The closing “If that’s what it means to be a Luddite, count me in!” turns the insult into an argument for asking questions (2018-12-15). - Preaching. Even when he draws on the Beatitudes, he offers them as “One intriguing starting point” that “require some re-interpretation” for a secular guide (2019-03-31). - Renouncing technology from privilege (2019-03-06).

Self-deprecation and admitted limits keep him on the reader’s level. - “Not being a historian, I wouldn’t know …” (2019-03-31) - “if I’m honest, there isn’t that much overt technological self-reflection in Endgame” (2019-04-26) - “many people disagree with me!” (2019-02-24) - “I must confess” (2019-03-06; 2019-03-31)

Changes of mind. There are no explicit reversals in this batch. What shows is concepts being built up in public: orphan risk gains new uses and a reciprocal twist, and the ethics critique sharpens from “good intentions alone are rarely enough” (January) to a full “ethics reset” (April). Republishing the 2015 Musk piece in December 2018 suggests he still held that view three years later.

6. What is distinctive#

1. One mind holding toxicology rigour and film criticism. In the same fortnight (5–7 March 2019) he publishes: - a 4,300-word case for dose-response thinking about algorithms - the closing chapter of a book arguing that sci-fi helps because it is “not tethered to scientific accuracy” - an op-ed on popcorn as a route to responsible innovation

He treats these as complementary ways of seeing, not separate careers. Few people in AI discussion move between Cochrane-grade evidence standards and Ex Machina with equal seriousness.

2. Exposure science carried over to AI. Most AI-ethics voices come from computer science, law or philosophy. His “algorithmic exposure”, meaning anyone affected by an algorithm’s decisions and not only its users, and his “exposure-response relationship” come from occupational and chemical risk. They put affected people at the centre and ask how a hazard actually reaches them. He also keeps hazard and risk distinct, a distinction the AI debate often blurs.

3. The early location of AI risk in human cognitive vulnerability. He locates it there rather than in superintelligence. We should worry more about “AIs that learn how to use our cognitive vulnerabilities against us” and need “tests that indicate when we are being played by machines” (FFTF p. 177, excerpted 2018-11-25; “press our metaphorical buttons”, 2019-06-12). He also treats superintelligence advocacy as a matter of belief, not evidence (“a religious convention”).

4. Risk framed as a threat to value and navigated as a landscape between intention and benefit. This lets him talk about dignity, autonomy and trust as risks without giving up quantitative risk science. He says it “extends conventional thinking rather than replacing it.”

5. Rejecting the “exceptional individuals” story about technology. “New technologies are developed by large teams of people, and their uses and abuses are tacitly sanctioned by millions more” (2019-04-26). Responsibility is spread across many people, and so is the agency to do better. This cuts against both founder hero-worship and the lone-villain framing common in AI discussion.

6. An early, non-polemical critique of OpenAI’s founding logic. In 2015 he argued that answering AI fears by accelerating AI “still adheres to the belief that the answer to technology innovation is… more technology innovation” (2018-12-15). He made this critique while defending Musk against the Luddite label.

7. Moral resources from outside the usual tech vocabulary. He takes design principles from the Beatitudes (2019-03-31): “Listen to those who are quiet”, “Elevate people who mend rifts and build bridges”. This is unusual in technology governance writing, and it matches his emphasis on bridge-building.

8. Humour as critique. The anti-guide and the bingo card use satire to make a serious point about expertise and exclusion through jargon. Very few risk scholars write this way.

7. The posts that best reveal how he thinks#

1. 2019-03-05 should-we-be-treating-algorithms. Structural analogy with stated limits, new concepts built by translation, a plausibility test against real systems, fair reading of evidence, tensions held (zero risk and acceptable risk, bias and human drivers), and quantitative risk science stretched by the value lens.
2. 2018-12-13 tech-startups-orphan-risks. Reframing through a gap in an existing taxonomy, and the founding statement of risk innovation, risk landscape and threat to value. It explicitly values that last concept for “opening up conversations”, and it shows the tool-and-business framing of the period.
3. 2019-03-06 navigating-the-risks-and-benefits (FFTF closing chapter). His self-image as an incomplete guide, “Don’t Panic”, obligation alongside responsibility, refusal of privilege and of deference to experts, why films work, and his admiration for creative recombination.
4. 2019-04-15 tech-companies-need-an-ethics-reset. Teaching as experiment, from values to operational outcomes. He is critical of the AI sector while assuming good faith, and threat-to-value reasoning appears in an ethics setting.
5. 2018-12-15 if-elon-musk-is-a-luddite. Refusing polarization, turning an insult into an argument, citing Late Lessons from Early Warnings as his own historical grounding, and making the early OpenAI critique.
6. 2019-03-31 design-principles-for-de-marginalizing. Humility about designing the future, “who decides” as the central question, and moral resources from an unexpected source.

(Close runner-up: 2019-04-26 avengers-endgame, for how he uses a blockbuster to think and also criticises the film’s own assumptions. For play, read 2019-02-08 how-to-give-the-best-scientific-presentation.)