B14 digest: 2024-02-11 to 2024-04-14#

What this batch is#

Nine weeks in early 2024: Sora, Gemini’s image failure, Neuralink’s first patient, a burned Waymo, Orchid’s embryo screening, the Apple Vision Pro. Maynard answers as Director of ASU’s Future of Being Human initiative and as a teacher. Six of the eighteen posts belong to an audio “In Conversation With …” series whose text is only show notes; they are set aside here. The rest is his own prose, apart from a student’s reflection in the multigenerational-class post and a later Modem Futura promo box in the Harari post.

AI is present but not dominant. Most posts are about technology in general, being human, and education, and AI is one of several transition technologies.

Main ideas#

1. Technology is part of what humans are, not an add-on (2024-03-31). This is the batch’s most important statement of his core outlook. He rejects the techno-optimist/pessimist question as being like asking whether he is an “oxygen pessimist or optimist”. Treating technology as “something we do and not something we are” is “misleading — and potentially dangerous”. He calls the navigation of advanced technology transitions (AI “and possibly AGI”, gene editing, biotech, quantum, neuroscience, nanoscale engineering and more, “synergistically accelerating”) one of humanity’s greatest challenges, with “catastrophic failures” possible. Three points stand out: - He criticises neighbouring fields (STS, ethics, responsible innovation, science policy, technology governance) as lacking “breadth of vision and the integration of understanding”. - He warns against assuming things will be fine “because it’s always done so in the past”. - He calls for “pilots”: new institutions and people able to steer these transitions.

2. Reading risk through orphan risks and value (2024-03-21). The Neuralink post brings back, and re-endorses, the “risk landscape” of orphan risks from his 2019 paper with Marissa Scragg. The method includes easy-to-miss risks, clustering and convergence, and threats to value across constituencies, with public backlash, trust, “bad actors” and organisational culture among its items. He says it “still stands up reasonably well”. This confirms that orphan risks is a Risk Innovation tool from at least 2019, applied here to neurotechnology rather than AI. He sets it beside two long-standing concerns: - Corporate power over bodies: “who “owns” you” when a company controls your implant, especially one from a sector where “monetizing the customer is the name of the game”. - An even-handed reading of the Silicon Valley “fail fast” playbook and its “reality distortion field”. He says it has worked, may speed up BCIs, and is something critics “discount … at their peril”.

3. What current AI is: counterfeit minds (2024-03-03). Using Dune, he names the AI he thinks we are actually building: not conscious or autonomous “thinking machines” but machines that “give the illusion of being able to do so — and in a very human way”. The risk he sees is the blurring of “human-ness”, not machine takeover. Rejecting technology is no refuge either: Dune’s machine-free society is “rife with political manipulation”. This links his 2018 framing of AI manipulation from outside the human club to later concerns about human-sounding AI.

4. Pragmatic AI governance, and a change of view on deepfakes (2024-02-25). He signs the deepfake open letter (criminalisation, developer liability), while saying he usually avoids such letters as “deeply naive” about governance. He had earlier declined the 2023 FLI pause letter. He explains the change: he once trusted people’s common sense to tell fake from real and is now “far less sure”, given deepfakes’ capacity to “destabilize relationships, communities, and whole societies”. For model failures such as Gemini, he prefers being “agile and responsive” after launch to demanding pre-launch perfection, and he credits Google and OpenAI for quick fixes. Generative AI is “neither robust, reliable, or trustworthy”.

5. Publics, backlash and the middle path (2024-02-18, 2024-04-14). Two posts show him refusing both reflexive opposition and uncritical boosterism: - On the Waymo burning, he attacks “lazy and dangerous” narratives of the public as unwilling “guinea pigs in a social experiment”, the othering of technology, and academics who implicitly legitimise vandalism. He warns that backlash could make development “more secretive, less open, and far less accountable”. - On embryo screening, he sets out a careful risk map: the burden of choice, “painful knowledge”, knowledge “once known isn’t unknowable”, disability and neurodivergence, commercial drift towards non-health traits (amplified by AI), a China race, equity, and diversity as species resilience. He calls it “a natural experiment … where we cannot put the genie back in the bottle”. He worries “A lot” but respects the founder and calls for “hard but respectful dialogue”.

6. Education as preparation for transitions (2024-03-10, 2024-03-17, 2024-03-24, 2024-02-11). - Against Harari, he proposes six durable areas: being human, flourishing, being part of something bigger, dynamic balance, creating opportunities, and building positive futures. The last centres on teaching “how to navigate advanced technology transitions”. - He adopts the playpen versus playground metaphor and is openly sceptical of edutainment and of technology as “a solution in search of a problem”. His rule for his own teaching is “the lowest level of tech necessary”. - In the Sora post he extends “playground of the mind” to young children. He imagines AI video as a language-driven “translator” of a child’s inner world, conditional on guardrails and a “substantial leap” in safety for children. An optimistic, formation-positive view that later work may complicate.

New or changed#

Deepfakes: an explicit change, from trusting common sense to supporting criminal law and developer liability. It is also a rare case of him signing an open letter.
Technology apologetics: he questions his decades-long habit of accommodating colleagues who see technology as separate from us. It is a shift towards more direct advocacy of a technology-centred view of human futures.
Spatial computing: within a week he is “less concerned” about a societal crisis. He calibrates with the literal precedent of laptops, tablets and smartphones.
Neuralink: more generous than in earlier posts. He credits the company with going slower, starting with medical uses and working with regulators, while moving his main worry to long-term patient autonomy and corporate dependency.
Responsible innovation: named among the fields he now finds too narrow, which is a mild distancing from a field he has long championed.

Concepts in this batch#

Oxygen analogy; technology apologetics; advanced technology transitions; pilots; orphan risks and risk landscapes; reality distortion field; who owns you (implants); counterfeit human minds; human-ness; agile post-launch correction; deepfake liability; othering technology; can versus should; ethics of choice; irreversible knowledge; natural experiment / genie in the bottle; diversity as resilience; six durable learning areas; playpens versus playgrounds; lowest level of tech necessary.

Most important posts#

2024-03-31 we-have-a-technology-problem-and: his core outlook (technology as constitutive; advanced technology transitions; pilots; the limits of existing fields).
2024-03-21 elon-musks-neuralink-plays-mind-games: orphan risks and risk landscape (2019, re-endorsed), the Silicon Valley playbook, corporate dependency.
2024-03-03 dune-part-two-artificial-intelligence: “counterfeit” human minds as his account of what current AI is.
2024-02-25 ai-rollercoaster-of-a-week: the change of view on deepfakes, scepticism about open letters, agile governance.
2024-04-14 welcome-to-the-age-of-swipe-and-select-embryos: a full worked example of his balanced ethics-and-risk method on a biotechnology.
2024-02-18 setting-fire-to-self-driving-cars-is-bad: his stance on public backlash and anti-technology narratives.