B06 digest (2021-03-28 to 2022-12-13, 13 posts)#

What this batch is#

This batch covers the period from the pandemic to the release of ChatGPT. It captures Maynard at a hinge point. His established identity as a nanomaterial risk scientist and a scholar of responsible innovation through science fiction is fully on show: graphene masks, advanced-materials standards, Jurassic Park. Alongside it, AI becomes a steadily larger subject, running through the neglect of AI risk (2021), Tesla Bot (2021), Alexa (2022), Midjourney (2022) and ChatGPT (December 2022). The batch has no Modem Futura podcast posts. Provenance issues are minor: ChatGPT answers (2022-12-08) and Midjourney images are excluded as evidence, and the chaos post is an excerpt from his sole-authored Films from the Future.

Main ideas#

1. Risk is about behaviour, exposure and process, not labels or hope. The two nanomaterial posts show the risk scientist’s core commitments that he carries into later work. - The graphene-mask post (2021-03-28) applies a four-question logic of exposure, hazard, mechanism and dose. It judges irresponsibility by process, not outcome. Selling an inhalation product without asking the obvious questions is “irresponsible innovation on a grand scale — even if the risks turn out to be negligible.” The burden of proof falls on manufacturers and regulators, and loss of public trust is counted as a harm. - The standards post (2022-02-10, originally delivered in 2020) separates terms of art from terms of science. It argues that governance built on definitional categories (“nanomaterial,” “advanced material”) is fragile, and that “nature doesn’t care what we call a material, it just cares about how it behaves.” It restates his 2010–11 “sophisticated materials” framework: five kinds of material that slip under the conventional risk radar. It also restates his 2011 Nature position against defining nanomaterials for regulation.

2. Risk as a threat to value, and orphan risks. The Tesla Bot post (2021-09-07) is the clearest statement in this batch of the ASU Risk Innovation approach. It asks three questions: what is important to you and the communities you touch; what orphan risks threaten it; and what small steps you can take. Orphan risks are defined as risks “hard to quantify and easy to ignore, and yet inevitably end up tripping innovators up if overlooked.” The concept already existed in 2021, linked to earlier Medium work on startups and investors, and sat inside a broader value-based view of risk. It is one tool among several here, not the organising idea. The 2021-08-03 post makes a related diagnosis of the field: AI ethics is booming while research on the risks AI itself presents is thin. Bibliometrically, most “AI and risk” papers use AI to manage other risks.

3. Innovation outpacing its consequences. The FRANKx lecture (2021-04-09) is his most ambitious conceptual piece in the batch. It sets out: - the solution problem: each solution seeds the next problem; - a claimed inversion of timescales: consequences now arrive faster than innovation can fix them, which means positive feedback, non-linearity, and the failure of trial-and-error and “simple models of innovation”; - base coding of the digital, the biological and the material, and transcoding between them (intelligent machines are a merging of all three); - bounded infinities and metaphorical quantum tunneling as ways to escape conventional thinking.

The conclusion is aimed at education: educate responsible “base coders,” develop tunnelers, and open learning beyond the walls of the elitist academy. The chaos-theory excerpt (2021-10-05) adds the butterfly effect, Perrow’s normal accidents and the Arkema plant: control is limited, but bounds separate plausible futures from fantasy.

4. Power and “whose future?” Across the Tesla Bot, Alexa and Jurassic World posts, his sharpest criticism is aimed at concentrated power to shape the future: - Musk and other wealthy “future-builders” imposing their visions (“innovators whose vision exceeds their understanding”); - Amazon driving users towards dependence as “super-consumers” under competitive pressure; - “power-hungry people” combined with powerful biotech.

Well-meaning scientific self-governance is praised, but will not be enough as capabilities spread. The remedy he proposes is broad participation in deciding “who gets to imagine the future.” He says little about specific regulation.

5. What AI is, and its risks (pre-ChatGPT to first ChatGPT). AI appears in several forms: - an embodied “brain” in a human-designed world (Tesla Bot); - predictive inference from fused data streams that may reach into health and mental states, and eventually brain-computer interfaces (Alexa); - a creative co-producer (Midjourney); - a fluent synthesiser and communicator (ChatGPT).

The risks he names are near-term and sociotechnical: reliability failures made worse by human behaviour; bias; privacy and inner-state inference; dependence and conditioning; attachment and gullibility; mis- and disinformation; hate speech; and the concentration of power. Robot-apocalypse framings are explicitly played down, and existential risk does not appear.

New or changed#

• AI moves from background to foreground. In 2021 AI is one “transcoding” product, or a risk-research gap. By late 2022 it is the subject of first-hand experiments.
• An enthusiastic first reaction to generative AI. The Midjourney and ChatGPT posts are strikingly positive. They talk of co-creation, unleashing creativity, AI explaining responsible innovation more clearly than experts, and a “leap in machine-augmented communication and engagement.” He also names the “serendipitous irony” of AI helping to overcome the challenges it creates. This sits in tension with his criticism of AI companies’ market logic (Amazon, Tesla) in the same period. Enthusiasm for the tool and wariness of the business model sit side by side. His concerns about gullibility and attachment in the ChatGPT post look like early signs of his later focus on cognition, formation and manipulation.
• Continuity rather than change on materials risk: the terms-of-art argument explicitly extends positions from 2005–2011. No reversal of view is signalled. The only self-correction is a sourcing regret (Bradbury) in the chaos addendum.
• Science fiction as a tool. Films reframe risk from spectacle towards responsibility and power; could-versus-should is his recurring touchstone.

Concepts appearing in this batch#

Irresponsible innovation (judged by process); the four-question exposure-hazard logic; “better safe than sorry”; terms of art versus terms of science; behaviour-based standards; sophisticated materials (five categories); Risk Innovation and risk as a threat to value; orphan risks; the gap between AI ethics and AI risk; could versus should; whose future and who gets to imagine it; the solution problem; the innovation-consequence timescale inversion; base coding and transcoding; bounded infinities; metaphorical quantum tunneling; the art of serendipity; the butterfly effect and bounded unpredictability; normal accidents; super-consumers and conditioning; human-AI co-creation and human-in-the-loop; machine-augmented communication; responsible innovation; public interest technology.

Most important posts for understanding his thinking#

1. 2021-04-09 bounded-infinities…: his big-picture theory of technology transitions (the solution problem, the timescale inversion, base coding and transcoding) and its implications for education.
2. 2021-09-07 should-we-be-worried-about-elon-musks-tesla-bot: Risk Innovation, orphan risks (2021 definition), layered AI risk, and the critique of tech leaders’ future-making power.
3. 2022-02-10 are-we-asking-the-right-standards-questions…: terms of art versus terms of science, and behaviour-based rather than definition-based risk governance, drawn from his nano work.
4. 2021-03-28 how-safe-are-graphene-based-face-masks: his applied risk-science method and the idea that innovation is irresponsible whatever the outcome.
5. 2022-12-08 i-asked-open-ais-chatgpt…: his first reaction to LLMs. It is optimistic about communication and names the risks of attachment, gullibility and disinformation.
6. 2021-08-03 we-need-to-get-more-innovative…: his diagnosis that AI ethics has crowded out AI risk research.