Quiz: Field Autopsy — Technology

Q: Minsky and Papert's *Perceptrons* (1969) demonstrated that: (a) Neural networks of any architecture are fundamentally flawed (b) Single-layer perceptrons could not compute certain functions (including XOR) — a correct result that was widely misinterpreted as applying to neural networks in general (c) Neural networks would never achieve practical results (d) Symbolic AI was the only viable approach to intelligence

(b) The mathematical result was correct but narrow. The authority cascade transformed it into a much broader claim — that neural networks in general were a dead end — which was an extrapolation, not a theorem.

Q: The neural network suppression (1969-2012) illustrates more failure modes operating simultaneously than any other case in the book. Which of the following was NOT a mechanism of suppression? (a) Authority cascade — Minsky's prestige caused the field to treat his extrapolation as proven (b) Consensus enforcement — funding agencies and conferences excluded neural network research (c) Democratic vote — AI researchers collectively voted to abandon the approach (d) Sunk cost — decades of symbolic AI investment created enormous switching costs

(c) There was no democratic vote. The suppression operated through structural mechanisms — prestige, funding, career incentives — not through collective deliberation. This is precisely the book's point: structural failure modes don't require conspiracy or deliberate action.

Q: "Capital-sustained error" refers to: (a) Errors caused by lack of funding (b) Wrong ideas that survive against the evidence because the volume of money invested creates an entire ecosystem with financial incentives to maintain the thesis (c) Errors in financial calculations (d) The high cost of scientific equipment

(b) This is tech's unique failure mode. The capital creates companies, jobs, conferences, and careers that depend on the thesis being correct — sustaining the idea through financial incentives rather than evidence.

Q: The three factors that enabled the neural network vindication in 2012 were: (a) New mathematical theory, more researchers, and better marketing (b) GPUs providing massive computation, internet-scale labeled datasets, and the dramatic AlexNet demonstration (c) Government funding, corporate interest, and media attention (d) Minsky's retirement, new textbooks, and policy changes

(b) The theory was available decades earlier. What changed was the infrastructure — hardware (GPUs), data (internet-scale datasets), and a public demonstration (AlexNet at ImageNet) that made the evidence undeniable.

Q: "Narrative-market fit" describes: (a) The alignment between a product and its market (b) The alignment between a technology story and what investors want to believe, which can sustain wrong ideas independently of evidence (c) The process of marketing a new technology (d) How startups find their target audience

(b) When narrative-market fit is strong, the technology doesn't need to work — it just needs to seem like it will work. The narrative attracts capital, capital creates activity, activity reinforces the narrative.

Q: The crypto ecosystem's defense against specific failures ("that wasn't real crypto") is structurally identical to: (a) The scientific method's hypothesis testing (b) The unfalsifiable defense pattern (Chapter 3) — attributing every failure to implementation rather than to the underlying thesis, just as strategic bombing advocates blamed "political constraints" rather than flawed doctrine (c) Legitimate quality control (d) A standard business pivot

(b) Both use the same unfalsifiable structure: no specific failure can disprove the core thesis because every failure is attributed to insufficient or incorrect application rather than to limitations of the approach itself.

Q: The technology sector's correction speed is: (a) Uniformly fast (b) Uniformly slow (c) Fast for technical questions (does the technology work?) and slow for narrative questions (is it good/transformative?) — because market feedback works for technical correction but not for narrative correction (d) Entirely driven by government regulation

(c) This asymmetry explains how the tech industry can simultaneously produce genuine breakthroughs and sustain massive narrative errors without experiencing cognitive dissonance. The two correction systems operate on different timescales.

Q: The autonomous vehicle timeline predictions illustrate a tech-specific error pattern of: (a) Insufficient engineering talent (b) Confusing rate of initial progress with rate of completion — early progress was rapid, but the last 10% of driving situations contains 90% of the difficulty (c) Deliberate deception by companies (d) Government regulation slowing development

(b) The difficulty curve of autonomous driving is not linear. Companies extrapolated from the easy part of the curve (highway driving, good weather, mapped areas) to predict completion, not recognizing that edge cases represent the vast majority of the remaining challenge.

Quiz: Field Autopsy — Technology

Q1. Minsky and Papert's Perceptrons (1969) demonstrated that:

(a) Neural networks of any architecture are fundamentally flawed (b) Single-layer perceptrons could not compute certain functions (including XOR) — a correct result that was widely misinterpreted as applying to neural networks in general (c) Neural networks would never achieve practical results (d) Symbolic AI was the only viable approach to intelligence

Answer

**(b)** The mathematical result was correct but narrow. The authority cascade transformed it into a much broader claim — that neural networks *in general* were a dead end — which was an extrapolation, not a theorem.

Q2. The neural network suppression (1969-2012) illustrates more failure modes operating simultaneously than any other case in the book. Which of the following was NOT a mechanism of suppression?

(a) Authority cascade — Minsky's prestige caused the field to treat his extrapolation as proven (b) Consensus enforcement — funding agencies and conferences excluded neural network research (c) Democratic vote — AI researchers collectively voted to abandon the approach (d) Sunk cost — decades of symbolic AI investment created enormous switching costs

Answer

**(c)** There was no democratic vote. The suppression operated through structural mechanisms — prestige, funding, career incentives — not through collective deliberation. This is precisely the book's point: structural failure modes don't require conspiracy or deliberate action.

Q3. "Capital-sustained error" refers to:

(a) Errors caused by lack of funding (b) Wrong ideas that survive against the evidence because the volume of money invested creates an entire ecosystem with financial incentives to maintain the thesis (c) Errors in financial calculations (d) The high cost of scientific equipment

Answer

**(b)** This is tech's unique failure mode. The capital creates companies, jobs, conferences, and careers that depend on the thesis being correct — sustaining the idea through financial incentives rather than evidence.

Q4. The three factors that enabled the neural network vindication in 2012 were:

(a) New mathematical theory, more researchers, and better marketing (b) GPUs providing massive computation, internet-scale labeled datasets, and the dramatic AlexNet demonstration (c) Government funding, corporate interest, and media attention (d) Minsky's retirement, new textbooks, and policy changes

Answer

**(b)** The theory was available decades earlier. What changed was the infrastructure — hardware (GPUs), data (internet-scale datasets), and a public demonstration (AlexNet at ImageNet) that made the evidence undeniable.

Q5. "Narrative-market fit" describes:

(a) The alignment between a product and its market (b) The alignment between a technology story and what investors want to believe, which can sustain wrong ideas independently of evidence (c) The process of marketing a new technology (d) How startups find their target audience

Answer

**(b)** When narrative-market fit is strong, the technology doesn't need to work — it just needs to seem like it *will* work. The narrative attracts capital, capital creates activity, activity reinforces the narrative.

Q6. The crypto ecosystem's defense against specific failures ("that wasn't real crypto") is structurally identical to:

(a) The scientific method's hypothesis testing (b) The unfalsifiable defense pattern (Chapter 3) — attributing every failure to implementation rather than to the underlying thesis, just as strategic bombing advocates blamed "political constraints" rather than flawed doctrine (c) Legitimate quality control (d) A standard business pivot

Answer

**(b)** Both use the same unfalsifiable structure: no specific failure can disprove the core thesis because every failure is attributed to insufficient or incorrect application rather than to limitations of the approach itself.

Q7. The "disruption myth" functions as tech's version of the revision myth because:

(a) It accurately describes how technology evolves (b) It smooths history by applying survivorship bias, converts systemic change into founder mythology, and pre-delegitimizes criticism by framing all resistance as self-interested defense of the status quo (c) It encourages healthy skepticism (d) It applies only to failed startups

Answer

**(b)** The disruption myth celebrates the technologies that succeeded and ignores those that failed (survivorship bias), attributes success to individual founders rather than structural conditions (hero narrative), and treats all criticism as evidence of the critic's obsolescence rather than as legitimate concern.

Q8. The technology sector's correction speed is:

(a) Uniformly fast (b) Uniformly slow (c) Fast for technical questions (does the technology work?) and slow for narrative questions (is it good/transformative?) — because market feedback works for technical correction but not for narrative correction (d) Entirely driven by government regulation

Answer

**(c)** This asymmetry explains how the tech industry can simultaneously produce genuine breakthroughs and sustain massive narrative errors without experiencing cognitive dissonance. The two correction systems operate on different timescales.

Q9. The autonomous vehicle timeline predictions illustrate a tech-specific error pattern of:

(a) Insufficient engineering talent (b) Confusing rate of initial progress with rate of completion — early progress was rapid, but the last 10% of driving situations contains 90% of the difficulty (c) Deliberate deception by companies (d) Government regulation slowing development

Answer

**(b)** The difficulty curve of autonomous driving is not linear. Companies extrapolated from the easy part of the curve (highway driving, good weather, mapped areas) to predict completion, not recognizing that edge cases represent the vast majority of the remaining challenge.

Q10. The neural network vindication primarily happened through:

(a) Persuasion — Minsky's successors were convinced by the evidence (b) Generational turnover — the old guard died off (c) Circumvention — external changes (hardware, data) enabled undeniable demonstrations that bypassed the old debate, consistent with Planck's principle (d) Government mandate — DARPA required neural network research

Answer

**(c)** The vindication came from outside the AI debate itself — GPUs and internet-scale data enabled demonstrations too dramatic to explain away. This is Planck's principle with a twist: the "new generation" didn't have to wait for the old to die; the evidence was undeniable enough to convert even some established researchers.

Scoring Guide

9-10 correct: Excellent. You understand both tech's unique failure modes and how universal failure modes operate within the technology sector.
7-8 correct: Good. Review the capital-sustained error concept (29.2) and the disruption myth (29.4).
5-6 correct: Fair. Revisit the neural network case study (29.1) and the correction speed analysis (29.5).
Below 5: Re-read the chapter with attention to how each failure mode from earlier chapters manifests specifically in the technology sector.