Michael Nielsen

We'll keep finding very fundamental new principles, analogous to Church-Turing or Noether's theorem, rather than exhausting the supply of deep ideas.

Quantum computers may handle a strictly larger class of interesting computations, and a quantum AGI would be qualitatively different from a classical one.

AI will help with data-intensive, well-specified problems (like protein structure) but won't automatically resolve the deeper bottlenecks needing long hostile verification loops or paradigm shifts.

The science-and-technology tree is so large that different civilizations explore radically different branches, implying large gains from trade between them indefinitely.

Progress outpaces raw falsification because the verification loop for a theory can take centuries; what drives it faster is aesthetic judgment and unification, which don't reduce to a formal procedure.

Naive falsificationism fails as a description of real science: facing an anomaly, scientists adjust auxiliary assumptions or suspect their instruments rather than drop the theory, and the right response is rarely obvious.

Discoveries need conceptual, technological and institutional scaffolding in place at once. Darwin needed deep time, voyage biogeography and fossil sequences; 'simple' ideas were gated by missing scaffolding, not missing intelligence.

The science-and-technology tree is far larger than we appreciate and we're near the bottom; diminishing-returns arguments fail because new fields keep opening and restoring low-hanging fruit.

Deep learning needs a demanding creative context with real stakes; time spent genuinely stuck is the most important part of internalization, and AI tools offer a seductive shortcut that substitutes for it.

Einstein later said he wasn't sure he was even aware of the Michelson-Morley paper when he developed special relativity, and that it wasn't dispositive for him.

Cosmic-ray muon experiments around 1940-41 confirmed time dilation: muons survive to ground level at rates explicable only if their time is dilated.

The Pioneer anomaly was eventually explained by asymmetric thermal radiation pushing the spacecraft, not by any modification of general relativity.

AlphaFold's success was largely a data-acquisition story: ~180,000 protein structures gathered over decades at a cost of several billion dollars; the AI was the final fitting step.