No. 298 of 500 No. 56 in Mind, Brain & Philosophy

The Myth of Artificial Intelligence

Why Computers Can't Think the Way We Do

By Erik J. Larson

Published
April 6, 2021
Print length
320 pages
Reading level
Intermediate
ISBN-13
978-0674983519
Publisher
Belknap Press: An Imprint of Harvard University Press
Written for
Readers weighing claims about artificial general intelligence who want a rigorous skeptical argument

Synopsis

Erik J. Larson, a computer scientist who spent years building natural language processing systems, argues that the belief in imminent human-level AI rests on a mistake about what intelligence is. He traces this intelligence error to Alan Turing and I. J. Good, who treated thinking as problem-solving that machines would eventually match, and follows it to modern futurists who present superintelligence as inevitable. Larson's central claim draws on the philosopher Charles Sanders Peirce, who distinguished three kinds of inference: deduction, which dominated early symbolic AI; induction, which drives machine learning; and abduction, the context-sensitive leap to the best available explanation that people make constantly in ordinary understanding.

Because no one knows how to program abduction, and combining deduction with induction does not yield it, Larson concludes that scaling current methods cannot produce general intelligence. He illustrates the gap with language-understanding puzzles that demand common sense and with the brittleness of data-driven systems, then criticizes big-data science for substituting volume for theory. His final argument is cultural: treating superintelligence as a foregone conclusion, he warns, devalues human insight and discourages the basic research a real breakthrough would need. The book appeared in 2021, before ChatGPT, so readers must judge how well its claims hold against today's language models.

Scorecard

Editorial score

How it scored

Editorial score: 62.8 / 1005 weighted criteria

  • Influence30%

    19.5/30
  • Authority25%

    17.5/25
  • Practical value20%

    11.0/20
  • Currency15%

    8.3/15
  • Readability10%

    6.5/10
  • Editorial score

    62.8/100

How the rank was computed

Rank No. 298 of 500 comes from a composite of 53.0: observable signals 24.7 of 55 points plus editorial review 28.3 of 45 points.

All 11 inputs
InputPoints
Observable signals24.7/55
Reach (Amazon ratings volume)9.5/16.5
Consensus across best-of lists8.3/13.8
Awards and bestseller evidence0.0/8.3
Reader rating, sample-adjusted6.0/8.3
Amazon sales rank0.9/5.5
Recency0.0/2.8
Editorial review28.3/45
Author authority7.9/11.3
Influence8.8/13.5
Practical value5.0/9.0
Readability2.9/4.5
Currency3.7/6.8

Each criterion is scored from 1 to 10 and weighted, and the points add up to the editorial score. The rank combines observable signals (55 percent) with this score (45 percent). Read the methodology.

Verdict

A sharp skeptic's case that data-driven AI lacks the abductive reasoning general intelligence requires, now tested by large language models.

Strengths

  • Clear explanation of deduction, induction and abduction
  • Grounded in practitioner experience with language systems
  • Strong historical account of AI's founding assumptions

Keep in mind

Published before ChatGPT; critics argue abduction is loosely defined and that Larson shows programming it is hard, not impossible.

Key takeaways

  1. Intelligence involves three kinds of inference, and today's AI handles deduction and induction while lacking abduction, the leap to plausible explanations.
  2. Successes in narrow tasks such as games and image labeling do not add up to general intelligence, however impressive each becomes.
  3. Common sense and causal understanding remain unsolved, and adding more data does not by itself close those gaps.
  4. Treating superintelligence as inevitable is a cultural myth that diverts funding and attention from basic research into how minds actually work.

Notable facts

About the author

Erik J. Larson

Erik J. Larson is a computer scientist and entrepreneur who has worked in AI since 2000, mainly on natural language processing. He worked at Cycorp, founded the DARPA-funded text-analysis company Knexient, co-founded Influence Networks, and in 2020 joined Knowledge Based Systems as a research scientist. He holds a philosophy PhD from the University of Texas at Austin and writes the Colligo newsletter.

Questions

What is The Myth of Artificial Intelligence about?
In The Myth of Artificial Intelligence, Erik J. Larson argues that the widespread belief in inevitable human-level AI is mistaken. Drawing on Charles Sanders Peirce, he says intelligence requires abduction, the ability to form plausible explanations from context, alongside deduction and induction. Because current AI relies on induction from data and nobody knows how to program abduction, he concludes that scaling today's methods will not produce general intelligence.
Has The Myth of Artificial Intelligence held up since ChatGPT?
Opinions divide. Large language models now perform well on many language-understanding tests long considered hard for AI, which critics cite against Larson's thesis. Supporters reply that these systems still make common-sense and causal errors and that no consensus exists on whether they perform genuine abduction. The book remains useful for its framework, but readers should test its claims against current evidence.
Who should read The Myth of Artificial Intelligence?
The Myth of Artificial Intelligence suits readers who want a careful skeptical counterweight to forecasts of imminent superintelligence, as well as students of philosophy of mind and cognitive science. It requires no programming knowledge, though the chapters on inference demand attention. Executives seeking guidance on deploying AI should look elsewhere, since the book addresses what AI cannot do rather than how to use it.

Sources

As an Amazon Associate, Best AI Books earns from qualifying purchases. Best AI Books is sponsored by Iternal.ai, whose founder and CEO, John Byron Hanby IV, wrote The AI Strategy Blueprint, our #1 ranked book. How we rank books.