In a 1995 New York Review of Books essay and the 2006 collection that took its title, Dyson articulated a view of science that emphasized dissent over consensus, curiosity over certainty, and the specific moral courage required to maintain minority positions against institutional pressure. The scientist, for Dyson, was not primarily a credentialed expert delivering authoritative conclusions but a rebel against whatever framework currently organized received knowledge. Science advanced through heresy, and the health of science depended on the institutional capacity to tolerate heretics. The Orange Pill cycle reads this framework into the AI moment: the silent middle that Segal describes is, in Dyson's vocabulary, the population of potential rebels currently being silenced by the discourse's demand for triumphalism or catastrophism. The capacity to hold both positions simultaneously — to dissent from both camps — is the specifically scientific disposition that the AI transition most urgently requires.
Dyson's framework was shaped by his own career as a productive heretic. He dissented from quantum field theory orthodoxy in the 1940s, from nuclear orthodoxy in the 1950s, from climate orthodoxy in the 2000s. Some of his dissents were vindicated; others were not. The point, for Dyson, was not that dissenters were always right but that the institutional capacity to tolerate them was the precondition for any future correction of current error.
The framework complicates Gramsci's hegemony in a productive way. Gramsci analyzed how dominant frameworks become common sense through institutional reinforcement. Dyson analyzed how dominant frameworks are eventually overturned — through the persistent work of dissenters who refuse to accept the common sense. Both analyses are necessary; neither is sufficient alone.
Applied to AI, the framework cuts in unexpected directions. The triumphalist discourse treats AI skeptics as Luddites standing against progress. The catastrophist discourse treats AI optimists as captured by Silicon Valley hegemony. Both discourses enforce consensus, and both foreclose the specifically scientific work of examining evidence without predetermined conclusions. The rebel position is neither consensus but continued investigation.
The framework also bears on the question of how AI should be governed. Institutions that enforce orthodoxy in safety discussions — whether the orthodoxy is 'accelerate at all costs' or 'pause everything' — suppress the very dissent that might identify problems current frameworks cannot see. The psychological safety that Edmondson identifies as essential to organizational learning is, in Dyson's framework, the institutional form of tolerance for rebellion.
The essay was originally a lecture delivered at the University of the Western Cape in 1992, published in The New York Review of Books in 1995, and expanded into the title essay of the 2006 collection. The South African context was significant: Dyson was speaking to a community that had recently emerged from apartheid and was thinking about what institutions of free inquiry would require.
Science as subversion. The fundamental scientific disposition is the refusal to accept authority — including the authority of current scientific consensus.
Institutional tolerance. Healthy scientific institutions tolerate dissent even when they cannot accept it; the capacity for future correction depends on preserving current heretics.
Minority positions. Progress comes from minorities willing to hold positions that majorities dismiss; the evaluation of such positions requires time and evidence, not vote counts.
The silent middle as potential rebels. The nuanced positions that the discourse suppresses are the positions from which eventual reframings emerge.
The framework has been criticized for romanticizing dissent. Not all heretics are right, and treating dissent as inherently valuable can license crankery alongside genuine insight. Dyson acknowledged this. His point was not that dissent is always correct but that institutions that cannot tolerate dissent cannot correct themselves. The distinction between productive heresy and mere contrarianism is a judgment call that communities must make case by case, and will sometimes get wrong.