The philosophical distinction between genuine novelty and mere recombination has a long history, running through Henri Bergson's élan vital, Alfred North Whitehead's creative advance, and Charles Sanders Peirce's tychism. What Smolin adds is a physical framework that makes the distinction rigorous rather than merely poetic. If the block universe is correct — if past, present, and future coexist as a determinate four-dimensional geometry — then genuine novelty is impossible. Everything that will happen is already implicit in what exists, and the apparent novelty of new configurations is an illusion produced by the limits of human perception. If, on the other hand, time is real and the future is open, then the resolution of configurations in the thick present can introduce outcomes that were not contained in the prior state.
The example most developed in Smolin's work is Stuart Kauffman's concept of the adjacent possible. At any moment, a system has access to configurations one step away from its current state. As the system explores these adjacencies, new configurations become accessible that were not previously adjacent. The space of the possible grows through the process of actualization. The specific molecules produced by early chemistry opened access to configurations that earlier chemistry could not have reached. The specific tools produced by early technology opened access to configurations that earlier technology could not have imagined. The space of what can be built expands through the process of building.
Applied to Dylan's 'Like a Rolling Stone' — the example Edo Segal develops in You On AI — the framework suggests that Dylan's creative process produced genuine novelty. Not because Dylan created from nothing (he didn't), but because the specific synthesis he achieved enlarged the space of what popular music could contain. Before 'Like a Rolling Stone,' certain configurations of lyrical, melodic, and structural possibility were not accessible. After it, they were. The song did not merely occupy an unoccupied region of an existing space. It expanded the space itself.
Can AI do this? The answer matters enormously for how to think about the current moment. Large language models explore vast spaces of possibility with speed and thoroughness no human can match. They produce outputs that are genuinely surprising. But the spaces they explore are defined by their training data and their architecture; the outputs are constrained by the processes that produced them. Finding surprising arrangements within an existing space is not the same as expanding the space. Whether future AI systems, with different architectures, could participate in genuine novelty is an open question. In the current paradigm, the distinction holds.
The concept of genuine novelty has roots in process philosophy (Whitehead, Bergson) and pragmatism (Peirce, James). Smolin's specific treatment is developed across Time Reborn (2013), Einstein's Unfinished Revolution (2019), and his collaborative work with Kauffman on combinatorial innovation.
Beyond recombination. Genuine novelty introduces something that was not implicit in the prior configuration — an enlargement of the space rather than a rearrangement within it.
Requires open future. Genuine novelty is possible only in a universe where time is real and the next state is not determined by the current state.
Operates in the thick present. The site where possibilities resolve into actualities that were not predetermined.
The adjacent possible. As configurations are actualized, new configurations become accessible — the possibility space grows through exploration.
The AI question. Whether current AI systems participate in genuine novelty or perform sophisticated recombination within fixed possibility spaces remains open.