Interview
Pamela McCorduck: Machines Who Think and the Early Days of AI | Lex Fridman Podcast #34
- Publication Context: Pamela McCordick's seminal 1979 work, Machines Who Think, was the first comprehensive history of artificial intelligence, created after McCordick, a novelist, decided to interview pioneers rather than write fiction about them.
- Funding Obstacles: McCordick secured funding for the project only after the National Science Foundation and other government agencies rejected her, with one official stating she was "only a writer, not a historian of science."
- Initial Support: The project was ultimately funded by a private grant from Ed Fredkin at MIT, who described McCordick's historical approach as a "crackpot idea" he was willing to support.
- Core Subject Matter: The book focuses on the four "founding fathers" of AI: John McCarthy, Marvin Minsky, Allen Newell, and Herbert Simon, all of whom were present at the 1956 Dartmouth Conference.
- Founder Motivations: Newell and Simon were cognitive psychologists aiming to simulate human cognition; Minsky viewed AI as a "cool" intellectual challenge; McCarthy approached it mathematically, believing it was a "great boon for human beings."
- Cultural Roots: McCordick argues AI stems from the ancient human impulse to "forge the gods," a tradition found in both Hellenic myths (where robots assist gods) and Hebraic law (the commandment against graven images).
- Religious and Psychological Fear: The resistance to AI often stems from a sense of blasphemy and a primal fear of being replaced, rather than just practical concerns like algorithmic bias.
- Literary Analysis of AI: McCordick challenges the literary convention of the "machine villain," citing Frankenstein to argue that the "monster" is created by human abandonment and lack of love, not the creation itself.
- Institutional Barriers: In 1974, the National Science Foundation initially sought to exclude AI from the definition of computer science, only including it after Don Knuth intervened.
- Expert Systems Era: The 1980s saw the rise of expert systems, which McCordick notes eventually hit a "wall" due to over-promising and the inability to scale symbolic processing to the complexity of human intuition.
- "AI Winter" Dispute: McCordick rejects the term "AI winter," characterizing it instead as a consequence of commercial hype cycles and over-promising, while basic scientific research continued unabated.
- Shift in Focus: The field has shifted from symbolic AI (logic and reasoning) to algorithmic and deep learning approaches, a transition McCordick views as having led to the embedding of human biases in modern decision-making systems.
- Santa Fe Institute Influence: A sabbatical at the Santa Fe Institute (1990–1992) provided McCordick with the vocabulary of "complex adaptive systems" to explain the work of painter and AI pioneer Harold Cohen.
- The "Geriatric Robot": McCordick describes a hypothetical AI companion designed to listen to the elderly, noting that human caregivers often frustrate patients, whereas a machine could offer optimized, unbiased companionship.
- Singularity Skepticism: McCordick is critical of the "Singularity" concept as popularized by Ray Kurzweil, arguing that machines already surpass humans in specific domains (like arithmetic) and that a sudden "game over" scenario is unlikely.
- Critique of Existential Threats: McCordick frames the fear of AI taking over (championed by figures like Elon Musk and Stephen Hawking) as the "male gaze," reflecting a patriarchal anxiety about being outcompeted rather than a logical assessment of AI capabilities.
- Women in Tech: Her 1996 book, The Future of Women, outlines four scenarios for women in technology, noting that despite the "Me Too" movement, women remain significantly underrepresented and "ground down" in the field.
- Feminist Motivation: McCordick reveals that her subconscious drive to support AI was partly an attempt to disprove the notion that intelligence resides exclusively in the "male cranium."
- Future Outlook: McCordick remains generally optimistic about AI but expresses concern about unpredictable "left field" events and the urgent need to imbue machines with ethics and empathy to match the broadened definition of intelligence.
- Prediction Philosophy: McCordick avoids making specific timeline predictions for AI breakthroughs, citing the example of deep learning (invented in 1986 but not applicable until decades later) to illustrate the non-linear nature of scientific progress.