Late Lessons, Jensen Huang and AI
Markdown version

B05 digest: August 2020 to February 2021#

What this batch is#

Thirteen posts from the COVID year, written while Maynard was associate dean in ASU’s College of Global Futures: - Cross-posts from the college’s Dean’s Blog, in an institutional, future-facing register. - Excerpts from Films from the Future. His 2018 book, republished with a fresh framing. - Material promoting Future Rising (October 2020).

All the text is his own. There is no AI-generated, guest or co-written material. The FFTF excerpts record what he thought in 2017–18, reaffirmed in 2020–21 by his choice to reissue them. Chronologically this batch comes before generative AI: “AI” here means machine learning, big data and prediction, not language models.

Main ideas#

1. Risk as a threat to value, and risk innovation. - Clearest in 2020-11-05 and the NASEM remarks in 2020-10-15. - Risk is unavoidable in a complex, causally tangled universe, where every action is “detrimental to someone in some way”. - Complexity and technological power multiply uncertainty, so “outmoded ideas about risk” become a risk in their own right. - He defines risk as “a threat to value”, with value taken broadly: economic, social and personal, including “equity, agency, and a sense of self”. - The main users he has in mind are entrepreneurs and innovators, not regulators. The Risk Innovation Nexus’s history (Michigan students → 2017 Accelerator → 2019 Nexus → end of seed funding in 2020) is spelled out.

2. Orphan risks, in their earlier form. - Orphan risks are already defined here: “hard to quantify threats to value that often slip between the cracks of conventional risk approaches”, or risks that are “easy to ignore” but “have a habit of coming back to bite”. - Two 2020 posts show the concept existed well before his 2026 work. It was built as an entrepreneurial and value-mapping tool and applied to Neuralink in a 2019 paper. - Here it is one tool within risk innovation, not a headline idea, justified systemically: non-linear systems “look as if they are thriving” until overlooked events trigger catastrophic failure.

3. Enhancement, identity and who owns you (brain-machine interfaces). The fullest risk analysis in the batch is about brain-machine interfaces (2020-08-28, 2020-10-15): - Identity is fragile and can be manipulated. - Connected brains will be hacked. - Human-made cyberspace carries our hubris (“hasn’t had billions of years of natural selection”). - Enhancement implants risk “the technological equivalent of indentured servitude” to the companies that supply them.

He makes a governance point that runs through his later work too: the real change is not new science but “a synergistic scaling of ability, accessibility, and use”, driven by Silicon Valley entrepreneurship that outpaces slow, careful institutional development and slips past the narrow risk frames of experts, here “medical neuroethics”. Experts who say “there’s nothing really new here” are “misguided”.

4. Prediction, bias and black-box machines. The Minority Report material (2020-09-26, repeated 2020-10-04) is his main AI-related critique in this period: - Phrenology → Lombroso → eugenics → fMRI → machine-learning “criminality” classifiers → Palantir-style predictive policing. - He treats this as a single structural lineage of seductive pseudo-prediction. - Science corrects itself, but “sometimes decades or centuries” late. - “Bad” behaviour is defined by social norms, not by morality. - Algorithmic bias builds prejudice into an “artificial judge and jury”. - We are building “artificial brains” whose workings we do not know.

5. AI risk: mundane but serious, not mainly existential. The one explicit AI-risk post (2020-11-12) names Musk, Hawking and Bostrom but moves the weight elsewhere: AI’s risks “are often far more mundane–but no less serious for this”, touching “autonomy, justice, equity, and our ability to have control over our lives”. AI sits “on a knife edge” between great benefit and catastrophe. The goal is benefit “for everyone, not just a privileged few”. Elsewhere he allows that machines “may one day surpass human intelligence” (2020-12-10), so he treats superintelligence as possible but not his focus.

6. Convergence and base code. The 2021-02-25 post restates FFTF’s idea of mastering “base code” (bits, DNA bases, atoms) and “cross-coding” between them, and argues this, not climate change, marks a real turning point in human history. The danger is irreversibility: “bricking” the world with no re-install option, and “tinker[ing] without understanding”. Climate change and pollution are recast as the results of “relatively crude technologies”, which shows how much harder base-code technologies will be. AI appears as an amplifier within convergence.

7. Responsibility to the future, trust and justice. - The Future Rising posts frame everything as a relationship with the future and a responsibility to it. - Humans are “architects of the future” whose abilities also “rob others of the futures they aspire to”. - He links climate change to “technological recklessness”. - On governance he endorses TIGTech (2020-12-15): trust is earned through trustworthiness; vaccine hesitancy should not be dismissed as irrational; and governance should be detached “from hype and ideology”.

8. Education and imagination. The Moviegoer’s Guide to the Future (2021-01-15) uses science fiction watched together as a creative, cross-disciplinary space for responsible innovation. He argues that “transformative learning has to be felt”, engaging the heart as well as the mind.

Concepts appearing#

Risk as a threat to broad value; risk innovation (and its mindset); orphan risks; outmoded risk ideas as risks; non-linear catastrophic failure; synergistic scaling of ability, accessibility and use; agile governance; technological indentured servitude; hacking and manipulation of connected minds; cyberspace as a human-made “fifth dimension”; “Here be monsters” mapping; the slippery slope of prediction; pre-justice; algorithmic bias; black-box “artificial brains”; slow scientific self-correction; mundane vs. existential AI risk; base code, cross-coding and “bricking”; tinkering without understanding; social base code; responsibility to the future; trustworthiness; entertainment-catalysed learning.

What is new or changed#

Most important posts in this batch#

  1. 2020-10-15 the-ethics-of-advanced-brain-machine-interfaces-and-why-they-matter: NASEM remarks combining risk innovation, orphan risks, entrepreneurial disruption, blinkered expert frames, and a change of view.
  2. 2020-09-26 the-seductive-slippery-slope-of-using-science-to-predict-bad-behavior: his fullest critique of machine-learning prediction, bias, opacity and the lineage of pseudoscience.
  3. 2020-11-05 risk-innovation-and-the-future: the risk concept in compact form, and the history of the Nexus.
  4. 2021-02-25 how-our-mastery-of-biological-physical-and-cyber-base-code-…: convergence, irreversibility, and the social extension of base code.
  5. 2020-08-28 navigating-the-complex-world-of-advanced-brain-machine-interfaces: identity, manipulation, hubris and corporate control of augmented bodies.
  6. 2020-11-12 is-artificial-intelligence-going-to-kill-us-all: short, but his explicit 2020 stance that AI risk is mundane but serious.