B25 notes: 2025-04-13 to 2025-05-25 (10 posts)#

Reading notes on Andrew Maynard’s Substack posts in batch B25. Only his own prose counts as evidence. Quotes are exact, including his original typos, curly punctuation and italics (shown as asterisks).

Batch context: spring 2025, the end of the academic year at ASU. Colossal Biosciences has just announced its “dire wolves”. The Trump administration has issued an Executive Order on AI education for K-12, a week after China announced plans to put AI into its school system. Agentic AI (reasoning models, Manus) is the hot topic, and Kasirzadeh and Gabriel have just posted a framework for classifying AI agents. Maynard is writing a responsible-AI module for a new undergraduate AI-literacy course, preparing to teach his Moviegoer’s Guide to the Future class for about the ninth time, marking two years of the Substack, and co-hosting the weekly Modem Futura podcast with Sean Leahy.

Three of the ten posts are Modem Futura episode posts. Following the user’s instruction not to spend time on the Modem Futura podcasts, they get one line each below and are not analysed, whatever their topic. The 2025-05-25 parasocial-communication essay mentions Modem Futura as the trigger for his thinking, but it is a stand-alone essay in his own prose, not an episode post, so it is covered in full.

Provenance problems in the rest of the batch are modest: - 2025-04-13 is mostly (about 6,900 of 8,000 words) a republished chapter of his sole-authored Films from the Future (2018). That is his own prose, but it records what he thought in 2017–18. The 2025 framing (about 1,100 words) is kept apart below. - 2025-04-24 attaches a Deep Research-generated comparison report (a PDF, not in the corpus). Only his bullet-point reflections in the post are used. - 2025-05-25 quotes ChatGPT o3, which he used as “reviewer 2”, in three footnotes. Those quotes are not his; his replies to them are. - 2025-05-11 includes the transcript of a course trailer. The narration is his script; the news clips it quotes are not his. - Perplexity “top takeaways” and ChatGPT “summary” buttons appear on several posts. They are links, not text in the post, and are ignored.

Relevance summary:

HIGH#

2025-05-18 — exploring-ai-through-cause-and-effect — “Exploring AI through cause-and-effect”#

Subtitle: “A new online tool that uses cause-and-effect relationships to explore and better-understand responsible AI development and use”.

Provenance. His own prose. It has two parts: a short introduction (about 330 words) and “an edited version of the script” (about 2,800 words) for two educational videos in a responsible-AI module for a new undergraduate AI-literacy course at ASU. The script was written “for a predominantly undergraduate audience” and “to be read aloud”, so it is simpler and more didactic than his usual essays. The tool itself (the Responsible AI Trajectories Tool, raitool.org) is his. The post does not say whether AI helped build it. Screenshots are of the tool; the header image is Midjourney.

Argument in his terms. - The teaching problem. AI moves so fast that “any set-in-stone course material would almost definitely be out of date before the first time it was taught”, and nobody wants a “talking head” lecture on AI ethics. So he aims at “fostering a responsible use mindset rather than simply focusing on facts and figures.” He wants a way of thinking that lasts, not content that dates. - Why responsible AI matters. AI is “arguably one of the most transformative technologies to have emerged over the past few centuries”. It could transform “nearly every aspect of our lives for the better” (personalised education, science, healthcare, resource management), “if we learn how to use it wisely”. But “as with all powerful technologies, there’s a danger of it causing serious harm”. He uses the Stan Lee line: with great power comes great responsibility. - Responsibility is everyone’s, and it is hard to define. He asks whether responsible AI is a matter for developers, for governments, or for “every user, every policymaker, and every person who interacts with AI”. His answer: everyone who uses AI or “whose life is impacted by it in some way, has a stake in” it. Defining “responsible AI” is “far from straightforward” because it depends on “context, values, goals, personal and cultural perspectives”. That makes one-size-fits-all principles hard. He lists the institutional responses (hundreds of ethics guidelines, industry safety research, regulation, OECD, Responsible AI UK, UNESCO) without judging them. He calls the whole thing “an extremely wicked problem”. In a footnote he insists on the strict sense: a problem where “attempted solutions alter the very nature of the problem”. - His proposed response is a culture of care plus cause-and-effect thinking. “One powerful and practical way” to steer AI toward good outcomes is “developing a culture and a mindset of care around AI development and use, and one of innovating responsibly”. Cause-and-effect thinking helps because we already know intuitively that actions have consequences. It sits “at the heart of much of the evidence-based decision making” in society, and “effects can be good as well as bad”. With AI it gets complicated because effects “grow, shift, and sometimes multiply, over time”. - The six models. Each is a curve in the tool with worked examples. 1. Linear: effect in proportion to cause (AI practice problems steadily improve grades). He calls it intuitive but “relatively rare when dealing with complex AI systems”. 2. S-curve: slow start, fast rise, plateau (investing in ethical AI design builds trust, then gains level off). It is reversible. 3. Exponential: feedback loops make effects multiply. He calls this “the doyenne of people who believe that AI is poised to transform society beyond all recognition — or lead to the collapse of society as we know it”. Such curves “rarely continue for extended periods” and become S-curves. But if the turn takes “years or decades”, they can still be “highly disruptive”, so spotting them early can “help prevent runaway impacts”. 4. Hysteresis: “a “memory” effect”. Effects are “sticky” and persist after the cause is withdrawn. His examples: a university that adopts output-maximising AI tools loses research depth and creativity, and the loss persists after the tools are discouraged; AI productivity tracking leaves a workplace “shaped by paranoia, burnout, and self-monitoring” after it is dialled back. Fixing unexpected consequences “often requires more than reversing an initial cause”. 5. Jagged: an S-curve with unpredictability on top: hidden feedback loops, external shocks, and uneven adoption, so “the risks and benefits are not evenly distributed”. His example is an AI sentencing tool used in some districts and not others, which leads to hidden biases, fluctuating decisions and “public trust becomes unstable”. Management lesson: “set it and forget it” fails; success needs “resilience, flexibility, and mechanisms for rapid course correction”. 6. Chaotic: predictable at first, then past a tipping point “small causes can lead to large, unpredictable, and irreversible effects”. His example is a person who forms “a deep emotional bond with an AI companion bot”, where “subtle behavioral shifts by the bot trigger unhealthy emotional dependencies, social withdrawal”, with effects that “persist long after the AI is removed”. At scale, chaotic AI impacts could destabilise “energy grids, financial networks, supply chains, or even societal cohesion”. These relationships “demand extreme caution”, because once chaos sets in “even well-intentioned interventions can worsen the situation”. What matters is early detection and prevention. - Effect as value and care. The conceptual move at the end: “when we talk about “effect” we are really talking about what we consider to be of value — or what we care for.” Effects can be money, but also access to education, wellbeing, or someone’s “ability to achieve what they aspire to”. So: “understanding effect as a threat to what is valuable to us and those around us — and what we care about, or what might actually enhance that value — becomes a powerful way of thinking about responsible AI.” Footnote 4 ties this directly to Risk Innovation (riskinnovation.org). Values vary: “wealth, health, and wellbeing” or “dignity, autonomy, and the ability to be part of building a better future”. The closing definition: responsible AI “is all about asking ourselves how our decisions or actions in the present will increase what we value in the future, or potentially jeopardize it”.

How firmly. Firm on the need for a responsible-use mindset, on responsibility being shared by everyone, and on non-linear, sticky and chaotic effects being the ones that matter. He openly admits the limits of his own frame: cause-and-effect thinking “is fraught with problems” and can feel “too cut and dried — a utilitarian approach … that ignores the messiness” of people, society and technology. He argues the tool’s nuances let that messiness “be recognized and approached with some nuance”. The models are “just six possible ways”, not a complete set.

Concepts and frameworks. - Responsible AI Trajectories Tool (raitool.org): six cause–effect models (linear, S-curve, exponential, hysteresis, jagged, chaotic) used as heuristics for anticipating the consequences of an AI use case. - Hysteresis: irreversibility and “memory” in the social effects of AI. This is a physicist’s concept put to social use. - Tipping points, feedback loops, runaway impacts: complexity-science language, continuous with his chaos-theory and Jurassic Park material. - Wicked problem (strict sense). - Culture / mindset of care: paired with “innovating responsibly”. This continues his March 2025 “hard” concept of care (2025-03-09 the-hard-concept-of-care-in-technology-innovation, B22), now applied in teaching responsible AI. - Effect = threat to (or enhancement of) value: risk innovation’s “risk as a threat to value”, recast as “effect” and joined to “what we care for”. - Responsibility as shared by all stakeholders (developers, businesses, governments, users, and those affected).

Analogies and comparisons. His phrase “as with all powerful technologies” is a generic structural comparison; no specific past technology is named. The critique of “going fast and not worrying if we break things” is a clear allusion to Silicon Valley’s “move fast and break things”. The curves come from physics and systems science and are used conceptually.

Views on AI. A transformative, general-purpose technology already “impacting nearly every aspect of our lives”, with influence that will grow “in ways we can’t fully predict yet”. The AI–human relationship “is nothing if not complex”, so AI’s effects are emergent, non-linear and path-dependent, not direct results of a tool’s design.

Views on AI risk. Harms are framed as threats to what people value, across scales: - individual: dependency on companion bots, social withdrawal; - institutional: loss of research depth, workplace paranoia, biased sentencing, unstable public trust; - systemic: destabilised critical infrastructure and “societal cohesion”.

Irreversibility (hysteresis, chaos) is the feature he stresses most. He hopes chaotic outcomes are rare at large scale (“Hopefully it’s one that we won’t see that often”). He treats the exponential/transformation narrative with some scepticism (such curves “rarely continue”), without dismissing disruption.

AI companies and leaders. Neutral here. Leading AI companies are “investing heavily in research on AI safety”. The “move fast” jab is the only critical note.

Governance and who decides. Everyone has a stake. He favours cultivating judgement and care in individuals and organisations over rules, since fixed principles fail across contexts. Adaptive management (“rapid course correction”, “early detection and proactive preventative measures”) is preferred to “set it and forget it”.

Cognition, language and formation. The education purpose is formation of a mindset (“fostering a responsible use mindset”) rather than transfer of facts. The companion-bot example points to emotional dependency as a serious harm. The hysteresis example about research depth and creativity implies AI can reshape intellectual habits in ways that outlast the tool.

Criticises / engages. Static course content; “talking head” ethics lectures; “move fast and break things”; “set it and forget it” management. Engages OECD, Responsible AI UK and UNESCO, and his own Risk Innovation work.

Change of view. No reversal. The notable development is that risk innovation’s value framing is now carried into AI literacy teaching and joined to the language of care he began developing in March 2025. He also uses the AI companion bot as his paradigm case of chaotic, irreversible harm.

Quotes. - “fostering a responsible use mindset rather than simply focusing on facts and figures.” - “understanding effect as a threat to what is valuable to us and those around us” - “lest in the name of going fast and not worrying if we break things, we find we can’t fix what ends up getting broken.” - “And the relationship between AI and humans is nothing if not complex.”

2025-05-04 — an-important-new-model-for-guiding-agentic-ai-oversight — “A new framework for guiding AI agent oversight”#

Subtitle: “As AI agents become more prevalent and powerful, a new paper outlines a framework for better-understanding how to govern and regulate them.”

Provenance. His own prose. It summarises and assesses Atoosa Kasirzadeh (Carnegie Mellon) and Iason Gabriel (Google DeepMind), “Characterizing AI Agents for Alignment and Governance” (arXiv 2504.21848). Descriptions of the four dimensions paraphrase the paper; short phrases inside quotation marks (e.g. “significantly reshape its environment across multiple dimensions, approaching full environmental control”) are the paper’s. The companion web app fvture.net/agentPlot was built “with AI’s help, of course” and posted “with a green light from Kasirzadeh and Gabriel”. The header image is Midjourney v7.

Argument in his terms. - Agentic AI is arriving, though the claims are hyped. He quotes Sam Altman’s January 2025 “Reflections” prediction that AI agents may “join the workforce” in 2025. Much of it “still reads more as hype than reality”. Still, interest in AI systems “that can make their own decisions and influence the environment they operate in” has sped up: reasoning models increasingly “decide how they’re going to solve problems”, and Manus is “shaking up what’s possible when an AI is given access to compute and the internet, and the license to interpret what it’s being asked to do”. He speaks of “an agentic AI wave” reaching research, teaching, business, healthcare and government. - This is genuinely new territory. “we’ve never had the ability to create machines that can decide on their own how to solve problems, and then — without human supervision — begin to alter the world around them to do this.” So governing systems whose self-chosen actions have “very material consequences” is increasingly important. - The first obstacle is framing, not rules. There is “no current accepted and easy-to-apply definition of an AI agent that is designed to support informed oversight and governance”. So “we’re not even sure yet how to formulate the problem of AI agent governance, never mind work out how to address it.” That is why he calls the paper “an important” step, though “just one step”. - His own working definition of an agent. It aligns with Russell and Norvig’s 1995 definition (perceiving and acting on an environment): “an AI that is able to determine how to achieve a set of goals — and adjust those goals if necessary — by manipulating the environment around it, whether this is digital, physical, environmental, behavioral, social, or a combination of these.” Footnote 5 says the domains are his, overlapping and possibly incomplete. Note that “behavioral” and “social” environments are part of his definition: an agent may act on people. - The four-dimension framework (autonomy, efficacy, goal complexity, generality), each with levels 0–5, modelled on SAE driving-automation levels: - Autonomy: A0 (human in full control) to A5 (no human oversight). - Efficacy: causal impact combined with type of environment (simulated, mediated through humans, physical). He values it because it separates impact from autonomy: a fully autonomous AI can have little impact, and a partly autonomous one can have “a profound and intentional impact”. - Goal complexity: the ability to break complex goals into sequenced sub-goals and replan. It is assumed to raise risk, and he sees it as increasingly important given Manus. - Generality: narrow to human-level adaptive competence across domains. He calls this “the “human dimension””, because human problem-solving in new conditions makes us “both incredibly effective at what we do and potentially very dangerous indeed”. - His critique of efficacy. “I’m not fully convinced that the proposed efficacy scale has it right yet”. It is unclear how the risks of an AI in a simulated environment (the “safest” type) might be realised, “or how direct causal effects on the beliefs, understanding, and behaviors of individuals and groups fits within the model.” The framework has a blind spot around epistemic and behavioural influence on people. That is one of his standing concerns (see the manipulation and cognition threads). - Why it matters for governance. Classifying agents along several dimensions allows governance that neither stifles “innovation because they treat agentic AI as a one-size-fits-all technology” nor fails to ensure responsible use “because they are designed for technologies that don’t match reality”. This is his usual both-ways framing: avoid over- and under-regulation by fitting oversight to the technology’s actual profile. - His own extension. The agentPlot radar app rotates the plot 45 degrees and lets the scale be “stretched”, “based on the assumption that the risk associated with going from 0 - 5 on each axis is likely closer to exponential than linear.” He plots a hypothetical embodied agent in a humanoid robot as an example. - Urgency. AI agents “are evolving rapidly, and (at the moment) are outpacing our understanding how to develop and use them responsibly.” He wants more work “— and fast —”.

How firmly. Firm that agentic AI is new and important, and that definition and characterisation must come before governance. He clearly endorses the framework as a necessary step, with a specific, hedged reservation about efficacy. The exponential-risk scaling is labelled an “assumption”.

Concepts and frameworks. Agentic AI / AI agent (his definition above); the Kasirzadeh–Gabriel four dimensions; the SAE L0–L5 levels as template; the exponential scaling of risk with capability level; the “human dimension” (generality); formulating the governance problem before solving it; environment types (simulated, mediated, physical), to which he wants to add direct epistemic and behavioural effects.

Analogies and comparisons. Autonomous vehicles (SAE levels): a structural template, and one the authors use explicitly. Waymo is one of the four example agents. Humans as the benchmark for generality and for danger. No comparisons with chemicals, nanomaterials and so on.

Views on AI. AI agents are machines that “decide on their own how to solve problems” and act on the world without supervision. That is new in kind. He is impressed by reasoning models’ “simulated reasoning” (footnote 1: they are read-only on the internet). He sees Manus-style agents as forerunners of a wave.

Views on AI risk. Risk rises with autonomy, efficacy, goal complexity and generality, probably exponentially. Physical-world efficacy and general capability are the main amplifiers. Under-specified risks include influence on beliefs and behaviour, and harms reached through simulated or mediated environments. He does not discuss existential risk.

AI companies and leaders. Altman is treated as a hype-prone forecaster. Researchers from Google DeepMind are treated respectfully as governance scholars. OpenAI, Anthropic and Google are mentioned for their reasoning models without evaluation.

Governance and who decides. Oversight should be fitted to agent profiles. No specific regulator or law is named. It is a researcher-led, framework-first approach.

Cognition, language and formation. Only through the efficacy critique: AI agents’ direct causal effects on “beliefs, understanding, and behaviors” need a place in governance frameworks.

Change of view. This continues his March 2025 Manus post (linked). His working definition of an AI agent is stated explicitly for the first time in this batch.

Quotes. - “we’ve never had the ability to create machines that can decide on their own how to solve problems, and then — without human supervision — begin to alter the world around them to do this.” - “we’re not even sure yet how to formulate the problem of AI agent governance, never mind work out how to address it.” - “how direct causal effects on the beliefs, understanding, and behaviors of individuals and groups fits within the model.” - “the risk associated with going from 0 - 5 on each axis is likely closer to exponential than linear.”

2025-05-25 — why-parasocial-communication-is-important — “Why parasocial communication around complex ideas is important – and why we need more of it”#

Subtitle: “Conveying facts and figures is an important communication strategy – but relationship-based communication and engagement have a unique and often under-appreciated role in empowering audiences at scale”.

Provenance. His own essay (about 2,400 words of main text plus about 1,100 words of footnotes). Not his: ChatGPT o3 quotes in footnotes 7, 9 and 10, where he used it as a “reviewer 2” and then answered it; a block quote from Horton and Wohl (1956) in footnote 2. His responses to ChatGPT are his own and useful. The header image is Midjourney v7. Modem Futura is discussed as his example, but this is not an episode post.

Argument in his terms. - The deficit model misreads people. Many academics were surprised by Trump’s 2024 support and concluded that his supporters had a knowledge deficit. “the irony here is that the most glaring “deficit” in this scenario is the lack of understanding that some people have of how individuals and society work.” What they missed is how much relationships shape how people “think, behave, and make decisions”, including parasocial relationships, where people feel “a personal and intimate connection with a public figure”. Footnote 1: Trump “successfully forged meaningful parasocial relationships with key audiences, and these mattered more than facts”. Its closing line is sharp: “sense tends to reflect the world as we wish it was, not as it is.” - Parasocial relationships cut both ways. “history is replete with examples of how feelings of connection and meaning — and of being “seen” — have led to widespread social manipulation and control.” Yet they can also “elevate and empower individuals and communities if used smartly and responsibly — especially when it comes to navigating highly complex technology transitions in an equally complex world.” - Transactional versus relational communication. Transactional communication has its place. But where it rests on the “deficit model” it “has been repeatedly shown not to be effective” and is still “deeply engrained in how we teach and train engineers, scientists and technologists”. Two-way public engagement (citizen juries, consensus conferences, participatory technology assessment such as ECAST, scenario methods) works for issues “where there are no black and white answers”, but it is “messy and inefficient”, needs experts “to make themselves vulnerable”, and is “incredibly hard to scale”. - A distinction he is still working out. “Relationship-based” communication uses relationships to reach set outcomes (trust in science, the “right” health decisions). “Relational” communication is broader: it helps information flow so that participants benefit “on their own terms”. “My thinking is still evolving here”. Parasocial communication is a form of relational communication that could have “the reach of transactional forms … and the impact of relationship-based communication and engagement”. It is little studied: footnote 8 reports his Scopus searches, which found one record on expert–public parasocial communication. - Examples. Christian Drosten’s Das Coronavirus podcast (an outlier, helped by the pandemic), Carl Sagan (“famously sidelined by academia”), Bill Nye, John and Hank Green, Alice Roberts. Audiences’ parasocial bond “effectively give[s] them permission to talk about what they know”; even seemingly one-way communication becomes “far more nuanced and relational” through this permission-granting. - Four purposes of expert–public communication: “Instruction (filling a perceived or actual deficit), ego (self-aggrandizement or self-promotion), impact (directly or indirectly setting out to bring about change or push an agenda), or empowerment (providing others with access to information that they are able to utilize on their own terms).” Academia assumes instruction and impact, where the communicator is secondary. Relational and parasocial communication puts the communicator at the centre, so it gets read as ego and risks being “censored”. Institutions may discourage it or try to co-opt “hard-won parasocial relationships … for other uses”. - His own stake. “In many ways it’s what I’ve aspired to for much of my career”: communication “that focuses on empowering others through relationship building”. Modem Futura made him see it through this lens. He defends the podcast’s long cold opens, personal stories and picky guest choices as relationship-building that “make[s] no sense whatsoever to someone with a transactional communication mindset”. The topics matter “for what it means to be human in the times we’re living through”. - Conclusion. The flow of knowledge “from where they are concentrated to where they are needed” will not go “from a trickle to a flood without” more relational and parasocial expert–public communication. That needs “a recalibration” by experts and institutions.

How firmly. Firm on the failure of the deficit model and on the value of relational communication. Explicitly tentative on the relational / relationship-based distinction and on the evidence base: “much of the evidence we have to go on here is anecdotal rather than systematic”. He pushes back on ChatGPT’s criticisms but concedes “The point is actually well taken” on progress in valuing public scholarship. As chair of his university’s promotion and tenure committee, reviewing “over 100 cases a year”, he still judges it “an uphill struggle”.

Concepts and frameworks. Deficit model (critique); transactional vs relational vs relationship-based communication; parasocial communication and relationships (Horton and Wohl 1956); permission-granting by audiences; the four purposes (instruction, ego, impact, empowerment); empowerment “on their own terms”; communication at scale; public engagement methods (consensus conferences, participatory technology assessment).

Analogies and comparisons. Trump’s podcast appearances, as a literal case of parasocial influence. COVID science communication (Drosten). Historical mass manipulation, noted without examples. There is no comparison with AI; he does not connect parasocial relationships to AI companions or chatbots in this post.

Views on AI. Not the subject. His use of ChatGPT o3 as “reviewer 2” is a practice worth noting. He found it stern, occasionally useful, and poorly informed about academic realities: “Clearly ChatGPT has not spent much time on the ground in academia!”

Risk, governance, expertise. This is his theory of how expertise should reach publics during “highly complex technology transitions”: relational, empowering, respectful of people’s own goals, and wary of both the deficit model and of manipulation. It underpins his public-engagement stance in risk and technology governance. He names the dark side (manipulation, control) as the same mechanism used badly.

Criticises / engages. Academics who hold the deficit model; academic cultures that treat public profile as vanity; institutions that co-opt relationships. Engages the science-communication literature (NASEM 2017 Communicating Science Effectively; Scheufele 2022; Simis et al. 2016; Staus, Risien and Cho), his PhD student Dania Wright’s dissertation on NSF Broader Impacts (“a deep chasm between intent and practice”), and ChatGPT as reviewer.

Change of view. Signalled: “I’ve only recently begin to think of this through the lens of parasocial communication.” The underlying commitment (empowering, relational engagement) is long-standing. The parasocial framing is new.

Quotes. - “the most glaring “deficit” in this scenario is the lack of understanding that some people have of how individuals and society work.” - “history is replete with examples of how feelings of connection and meaning — and of being “seen” — have led to widespread social manipulation and control.” - “especially when it comes to navigating highly complex technology transitions in an equally complex world.” - “sense tends to reflect the world as we wish it was, not as it is.”

2025-04-13 — de-extinction-conservation-futures — “Ancient wolves, conservation futures, and one of the fastest growing biotech startups in history”#

Subtitle: “Colossal Biosciences has been grabbing headlines with claims of bringing the long-extinct dire wolf back from the dead. But how responsible is the company’s vision of “commercial evolution?”“

Provenance. All his own prose, from two periods. - 2025 framing (about 1,100 words, plus footnotes 1–2): Colossal Biosciences, the dire wolves, the “wooly mouse”, “commercial evolution”. - Republished chapter (about 6,900 words): Chapter 2 of Films from the Future (Mango, November 2018), “Jurassic Park: The Rise of Resurrection Biology”, with its original endnotes. It was written around 2017; it refers to Hurricane Harvey “as I’m writing this”. Footnote 9 carries a 2025 aside: “remember this was written in 2018 — things have progressed since then”. Earlier batches already cover this chapter (B02 2018-08-26; the chaos section in B06 2021-10-05; audio show notes in B09). Republishing it in 2025 shows he still stands by it. - Not his: quotations from Colossal’s website; Ian Malcolm’s lines; the Buzzfeed logbook quote. The header image was “reimagined by ChatGPT 4o”. The Perplexity and ChatGPT buttons are links only.

Argument in his terms (2025 framing). - Colossal as a direct outgrowth of what he wrote about in 2018. He treats it as “one of the fastest growing biotech startups around” ($10.2 billion valuation). The dire-wolf “rebirthing” is “a bit of a misnomer”: the animals are “at best, genetically altered gray wolves”. Still, the precision gene editing is “impressive”. - Hubris. Colossal wants “smart humans to step up to the plate and show nature a thing or two.” “The hubris here is palpable.” Many scientists dispute that de-extinction answers mass extinction or that the technology is as transformative and ethical as claimed. He notes the funding: venture capital including In-Q-Tel, “the venture capital arm of the CIA”. He also notes the Apollo analogy Colossal uses for “commercial evolution”. - But he stays open about the technology. “as precision gene editing continues to merge with advanced AI capabilities that could open up pathways to massive multi-gene editing approaches, Colossal may be onto something — at least technologically.” It “may be that this is the technological future of conservation biology — not adhering to a nostalgic natural selection-centric past, but embracing a “commercial evolution”-driven future.” This is his characteristic even-handedness: he criticises the hubris without rejecting the technological trajectory. - The central question is responsibility. As the ability to re-engineer nature grows “and the ambitions driving it continue to balloon — so do the questions around what responsibility means in an era of de-extinction.” - AI companies as a reference point for self-regard. Footnote 1: “I thought that AI companies like OpenAI, Anthropic and others were full of themselves, but Colossal takes this to a whole new level. You have to admire their audacity though!” - AI–biotech convergence. Footnote 2 links his Evo 2 post (“An AI model that can decode and design living organisms”). AI is presented as an accelerant of genetic design.

Argument (2018 chapter, reaffirmed). - Jurassic Park is about “greed, ambition, genetic engineering, and human folly”. Beyond the technology, it “reveals a very human side of science and technology”, including “the sometimes oversized roles mega-entrepreneurs play in dictating how new tech is used, and possibly abused.” - Beyond de-extinction: re-imagining species; synthetic DNA with extra bases; Venter’s JCVI-syn3.0; “a possible transition from biological evolution to biology by design”; genetic watermarking and ownership (“who owns them?”). - Could vs should. He has met “remarkably few scientists and engineers who would consider themselves to be unethical or irresponsible”, but many are so absorbed in their work that they “struggle to appreciate the broader context”. The lysine contingency and the all-female dinosaurs are “a salutary tale of scientists who are trying to be responsible—at least their version of “responsible”—but are tripped up by what they don’t know, and what they don’t care to find out.” Responsible science “is about more than just having good intentions”. It means “having the humility to recognize your limitations, and the willingness to listen to and work with others who bring different types of expertise”. “the more complex the science and technology … the more pressing this distinction between “could” and “should” becomes.” - Past technologies as “rehearsal”. The Industrial Revolution and the atomic bomb were cases of balancing what we can and should do. Those challenges “were just a rehearsal for what’s coming down the pike”. - Complexity. Chaos theory: the butterfly effect, bounded unpredictability, “points of stability”. The Arkema organic-peroxide plant during Hurricane Harvey (overflowing toilets, a snake) is a real cascade in the chemical industry. He cites Perrow’s “normal accidents”. “we cannot wield perfect control over complex technologies within a complex world”. - Power. The film is a morality tale about corporate greed, but that reading is “too simplistic”. Power differentials are inevitable and often legitimate (including “the fiduciary responsibility of innovators to investors”) and they stop society stagnating. “The challenge we face is not to abdicate power, but to develop ways of understanding and using it in ways that are socially responsible.” Scientists, activists, legislators and citizens all wield power. He acknowledges “the past two hundred years of environmental harm and human disease tied to technological innovation” and also the good that profit-driven innovation has done.

How firmly. The 2025 framing is openly ambivalent: sharp on hubris, open on the technology. The chapter is firm on could/should, humility, complexity and responsible use of power.

Concepts. Resurrection biology / de-extinction; commercial evolution (Colossal’s term, which he examines); “thoughtful disruptive conservation” (Colossal’s); biology by design; could vs should; responsible science as humility plus listening; normal accidents; bounded chaos and points of stability; responsible use of power; responsible innovation; corporate social responsibility; ethical entrepreneurs.

Analogies and comparisons. Gene editing and biotech are the subject (literal). The chemical industry (Arkema) is a literal case of cascading failure in a complex system. The Industrial Revolution and nuclear weapons are historical precedents, used structurally. AI companies are a structural comparison for corporate hubris. AI is also an enabling technology that converges with gene editing. Jurassic Park (the film) is a conceptual lens.

Views on AI. Here AI is an accelerant of biotech (AI-enabled multi-gene editing), and AI companies are a benchmark for grandiosity. There is no direct AI-risk claim.

Criticises / engages. Colossal Biosciences (Ben Lamm, George Church): hubris, grandiose rhetoric, Apollo analogy. Crichton’s chaos “hokum”. Hammond as mega-entrepreneur. Scientists’ myopia.

Change of view. None signalled. He reaffirms his 2018 analysis in the face of a real company.

Quotes. - “The hubris here is palpable.” - “I thought that AI companies like OpenAI, Anthropic and others were full of themselves, but Colossal takes this to a whole new level.” - “a salutary tale of scientists who are trying to be responsible—at least their version of “responsible”—but are tripped up by what they don’t know” (2018 text) - “The challenge we face is not to abdicate power, but to develop ways of understanding and using it in ways that are socially responsible.” (2018 text)

MEDIUM#

2025-04-24 — us-and-china-vie-for-ai-k-12-leadership — “US and China Vie for AI Leadership in K-12 Education”#

Subtitle: “Two new national plans promise to overhaul classrooms with AI. Here’s how the US and China differ – and overlap”.

Provenance. The post (about 900 words) is his own prose, including his “first-cut top-level reflections” in seven bold-headed points. An attached PDF comparison report (not in the corpus) was produced with OpenAI’s Deep Research. Footnote 1 says “usual cautions apply for AI-generated analysis” and describes a ““developed by AI, checked by humans” workflow” being developed in ASU’s Future of Being Human initiative. His reflections may draw on that report, but they are presented as his. The header image is Midjourney.

Argument in his terms. - The US Executive Order on Advancing Artificial Intelligence Education for American Youth (issued the day before) partly rebuilds “from a systematic dismantling of AI-focused initiatives from the previous administration” (the Biden 2023 EO, rescinded in January). It responds to awareness that the US “cannot afford not to invest” in AI skills, spurred by China’s “breakneck speed”. - Shared ground. Both see AI as transformative “(even if it doesn’t reach the hyperbolic heights of artificial general intelligence any time soon)”. Both see the geopolitical stakes, where not investing could mean “global irrelevance”. Both expect smart integration to “most likely transform learning in positive ways”. - Divergence. The US emphasises economic growth, geopolitical power and “personal empowerment”. China adds technological independence, “collective advancement and societal flourishing”, “soft power”, and “A-HERO” qualities (AI literacy, High-order thinking, Ethical thinking, Resilience, Openness). - Responsible AI. China places “a more overt emphasis on AI ethics, responsible use, and public value growth beyond economic growth”. In the US it is “implied” (his typo “Un the US”) that these will follow naturally from economic and geopolitical strength, “although this is not explicitly indicated”. This is a quiet criticism of the US order. - Centralised vs distributed governance. China’s centralised approach is “fast, agile” but has “built-in rigidity that may make it fragile in a rapidly changing world”. The US’s distributed, partnership-based approach risks “slower and more patchy adoption, potentially leading to students in some areas being left behind”. Its benefits are “far greater resilience under uncertainty as AI develops.” - Collaboration. “win-win opportunities” for US–China collaboration, as AI “transcends national politics”.

How firmly. Tentative: “first-cut” reflections on an order “only a day old”.

Concepts. Resilience vs fragility under uncertainty; centralised vs distributed governance; soft power; AI literacy; public value beyond economic growth; “developed by AI, checked by humans”.

Views on AI. Transformative but not AGI “any time soon”. This is a rare direct statement of scepticism about near-term AGI.

Governance. An even-handed comparison. He values resilience and adaptability under uncertainty (a complexity theme) and notes the equity risk (“left behind”) of distributed approaches. He faults the US order’s silence on ethics and responsibility. He favours international collaboration over zero-sum competition.

Education. K-12 AI integration is assumed to be necessary and likely positive if “smart”.

Quotes. - “even if it doesn’t reach the hyperbolic heights of artificial general intelligence any time soon” - “At the same time, the country has built-in rigidity that may make it fragile in a rapidly changing world.” - “The benefits are far greater resilience under uncertainty as AI develops.”

2025-05-11 — designing-responsible-technological-futures — “Designing the technological futures we aspire to”#

Subtitle: “Why transformative technology demands responsible innovation — and the critical questions we need to be asking before it’s too late”.

Provenance. A short post. About 300 words are his framing prose. The rest is a transcript of the trailer he made for ASU’s Moviegoer’s Guide to the Future class. He wrote and put together the narration (“shamelessly inspired by some work I did with a couple of producers”), so the bold NARRATOR lines count as his. The CLIP lines are news and film excerpts and are not his (e.g. “What if the biggest AI threat of all, is to human relationships.” is a news clip). The trailer seems to date from around 2024: the clips mention Neuralink’s first human implant and gene-editing approvals.

Argument in his terms. - He is preparing to teach the class “for the 9th time”. Rewatching the trailer, he finds its questions “more relevant than ever to the questions all of us should probably be asking about transformative technologies, from AI and gene editing to brain computer interfaces cloning, and much more.” - The narration: technologies are “advancing at a thousand miles an hour”. “we’re on the edge of being able to redesign the very essence of what it means to be human”. “too few people are asking where we’re going and how we get there”. Sci-fi movies “kickstart conversations on what it means to innovate responsibly, and ethically” and “open up pathways in our imagination to build the futures we want, while avoiding those we don’t.” - His closing restatement: “What are the technologies we want in our lives? Where do we want these technologies to take us? And how do we spot the dangers before it’s too late?” “We have an incredible responsibility to get this right. Otherwise, we risk losing everything in our pursuit of technologies we do not understand, and cannot handle.”

How firmly. Self-deprecating about depth (“it’s short, and it’s not that deep”), but he says the closing thoughts “should be on everyone’s mind”.

Concepts. Responsible and ethical innovation; science fiction as a way into imagining futures; futures we want vs futures we avoid; redesigning what it means to be human; anticipating dangers “before it’s too late”.

Analogies. AI grouped with gene editing, BCIs, cloning, de-extinction (the Tasmanian tiger clip) and xenotransplantation, as one class of “transformative technologies”. Science fiction film as conceptual lens.

Views on risk. Stark framing: the risk of “losing everything” through technologies “we do not understand, and cannot handle”. This is stronger rhetoric than his usual even-handedness, though it is trailer copy.

Quotes. - “What are the technologies we want in our lives? Where do we want these technologies to take us? And how do we spot the dangers before it’s too late?” - “Otherwise, we risk losing everything in our pursuit of technologies we do not understand, and cannot handle.”

LOW / NONE#

• 2025-04-15 — navigating-the-shift-to-electric-vehicles — “The Secret Lives of Electric Vehicle Owners” (none). A Modem Futura episode post with Jamey Wetmore (ASU; Tech Skeptic Goes Electric) and Sean Leahy on EV ownership, charging and data privacy. It has a short intro in his prose praising “informed skepticism”, plus ChatGPT o1-Pro-generated entry points. Skipped per user instruction.
• 2025-04-20 — surprised-by-serendipity — “Surprised by serendipity” (low). His prose. A two-year Substack anniversary roundup: subscriber numbers (over 4,500; 30–40k monthly views), his three aims for public writing (making research accessible; writing “as a discipled way to stay on top of new developments while developing and extend my own thinking”; honing communication), a “serendipity” button, Perplexity and ChatGPT buttons to “continue the conversation using AI”, and his Prompt2URL app, “spun up” with ChatGPT. He is amazed at “how fast and easy “vibe coding” with AI is”. Relevant only to ai-in-scholarship-writing.
• 2025-04-22 — beyond-the-heros-journey-ai-protopia — “Beyond the Hero’s Journey: AI, Protopia & the Future of Film” (low). A Modem Futura episode post with filmmaker and futurist Taryn O’Neill on AI-powered moviemaking and on how stories shape futures. It has a short intro in his prose plus ChatGPT o3-generated entry points. Skipped per user instruction.
• 2025-05-06 — what-students-really-think-about-ai — “What students really think about artificial intelligence” (low). A Modem Futura episode post with ASU sophomores Bella Faria and Caleb Lieberman (“ChatGPT-natives”). It has a longer intro and a section of “my key takeaways” (peer-to-peer learning, creative innovation, uneven AI use across classes, cheating, the value of a university education, the future). Topically relevant to education-higher-ed, but skipped per user instruction and not analysed.