The Tradition Did Not End With the Greeks
This book has focused on five ancient thinkers. But the questions they raised did not stay in Athens. They were carried forward by thinkers who faced their own versions of the same fundamental problems. Four modern figures deserve attention, because their work connects directly to the AI challenges explored in the preceding chapters.
Hannah Arendt and the Banality of Thoughtlessness
Hannah Arendt was a German-Jewish political philosopher who fled Nazi Germany and spent most of her career in the United States. She is best known for her coverage of the trial of Adolf Eichmann, the Nazi bureaucrat who coordinated the logistics of the Holocaust.
Her most disturbing finding was not that Eichmann was a monster. Her finding was that he was not. He was, in her famous phrase, terrifyingly ordinary — a man who had simply stopped thinking, who had replaced personal moral judgment with the execution of orders, who had found in bureaucratic process a comfortable substitute for the difficult work of ethical reflection.
She called this the banality of evil, but the more precise formulation is the banality of thoughtlessness. The danger was not malice. It was the abdication of judgment.
The connection to AI should be apparent. The risk of AI-augmented organisations is not that the AI is malicious. The risk is that the humans within those organisations stop thinking. The danger is not the machine. The danger is the human who stops thinking because the machine is there.
Michael Polanyi and Tacit Knowledge
Michael Polanyi was a Hungarian-British chemist and philosopher who developed the concept of tacit knowledge: the idea that we know more than we can tell. Skilled practitioners carry knowledge they cannot fully articulate — knowledge embedded in perception, in habit, in the trained sensitivity that comes from years of practice.
If there is a significant body of professional knowledge that is genuinely tacit — that cannot be articulated in language and therefore cannot be captured in training data — then there are categories of expertise that AI systems cannot acquire no matter how much data they are trained on, because the data does not contain the knowledge.
The experienced QA engineer who can sense that something is wrong with a codebase before they can say what it is. The project manager who senses that a client relationship is deteriorating before any metric reflects it. That knowledge is real, professionally valuable, and — in Polanyi's terms — fundamentally non-transferable to systems that learn from explicit data. It is Aristotle's phronesis, restated in the language of twentieth-century epistemology.
Daniel Kahneman and the Two Systems
Daniel Kahneman, the psychologist who won the Nobel Prize in Economics, spent his career documenting the systematic ways in which human judgment fails. His framework of System 1 (fast, intuitive, pattern-based) and System 2 (slow, deliberate, analytical) thinking maps directly onto the practical challenges of working with AI.
AI systems are extraordinary System 1 thinkers. They are fast, pattern-matching, fluent, and confident. What they are not — what the current generation of models cannot be, by design — is System 2 thinkers. They do not slow down. They do not notice when the pattern they are matching is subtly wrong for the situation at hand.
The human working alongside AI needs to supply the System 2 thinking. The slow deliberation. The willingness to question the fast, confident output. This is Socratic elenchus restated in the language of cognitive psychology. The names are different. The method is the same.
Ethan Mollick and the Partnership Imperative
Ethan Mollick, a professor at the Wharton School of Business and author of Co-Intelligence, has done more than almost any contemporary thinker to bridge ancient concerns about judgment and the immediate practical reality of working with AI systems today.
Mollick's core recommendation is elegantly simple: invite AI to the table, but never give it the head of the table. Use it for what it does extraordinarily well — generating options, synthesising information, accelerating execution — while retaining human authority over what matters: judgment, values, responsibility, and the final decision.
This is not a temporary compromise until AI gets good enough. It is a permanent structural principle grounded in a clear-eyed understanding of what each party brings. The AI brings speed, breadth, and pattern-matching at scale. The human brings stakes, context, accountability, and the practical wisdom to know when the pattern is wrong.
Mollick names this co-intelligence: not human intelligence versus artificial intelligence, but a collaborative mode in which each amplifies the other's strengths and compensates for the other's weaknesses. Socrates would have recognised the principle. He spent his career in dialogue — never asserting, always questioning, always finding that the best thinking emerged from the encounter between two minds rather than the monologue of one.
The examined machine is not a solo project. It is a conversation.
Practical exercises
- Identify one moment this week when you almost stopped thinking because the AI had already produced something plausible. What did it cost you to not slow down? What might it cost you next time?
- Write down one piece of knowledge you carry about your field that you have never seen successfully captured in a document or training dataset. That is your tacit knowledge. That is where your irreplaceability lives.
- Map your current AI use against Mollick's principle: where is AI at the table, where is it at the head of the table, and where is it not invited at all? Adjust accordingly.