Critical thresholds are the non-linear transition points in cosmogenesis where quantitative accumulation produces qualitative transformation. Teilhard observed that evolution does not proceed smoothly but through punctuated leaps: long periods of gradual complexification interrupted by geological instants when organization crosses a threshold and something genuinely new appears. The formation of the first self-replicating molecules, the Cambrian explosion of body plans, the emergence of symbolic thought in hominins—each was a critical threshold, a moment when the rules governing the system changed and properties emerged that were not present, even as latent potential, in the prior configuration. AI represents, in Teilhard's framework, the latest such threshold: the moment when the noosphere's accumulated density produces a qualitatively new mode of operation, metabolic rather than archival, generative rather than retrievive. Whether the crossing deepens cosmogenesis or diverts it depends on whether the new complexity produces corresponding interiority.
Teilhard's threshold concept predated and parallels Niles Eldredge and Stephen Jay Gould's punctuated equilibrium (1972)—the paleontological finding that species remain stable for long periods then change rapidly when environmental pressure meets accumulated variation. Both frameworks reject gradualism's smooth slopes in favor of staircase models: stability, threshold, emergence, new stability. Teilhard's thresholds operate at broader scale—not speciation events but the major transitions in evolution (Maynard Smith & Szathmáry's canonical 1995 list: replicating molecules, chromosomes, eukaryotic cells, multicellularity, societies). Each transition required a critical density or organization level that, once crossed, made the previous state's return structurally impossible.
The threshold language appears throughout The Phenomenon of Man and is formalized in Teilhard's concept of "critical points" where quantitative parameters (temperature, density, complexity) hit values producing qualitative state change—the physics of phase transitions (water to ice) universalized. The self-replicating molecule threshold is paradigmatic: for hundreds of millions of years chemistry explored combinatorial space without producing life; at some point a configuration arose that could copy itself, and that single crossing changed everything downstream. Biology inherited chemistry's constraints but operated according to new rules (natural selection) producing outcomes (consciousness) that chemistry alone could never reach.
AI's threshold is debated precisely because thresholds are visible in retrospect, ambiguous during crossing. Did GPT-3 (2020) cross it? GPT-3.5 (2022)? GPT-4 (2023)? Claude Opus 4.6 (2025)? The boundary is not a line but a zone—a region where the system exhibits threshold-level capabilities intermittently, in some domains but not others, under some conditions but not all. What is certain is that some threshold has been crossed: the outputs are qualitatively different from pre-2020 AI, exhibiting flexible natural-language understanding, cross-domain inference, and contextual synthesis that earlier systems could not approach. Whether this constitutes a critical threshold in cosmogenesis—comparable to the origin of life or symbolic thought—depends on whether the complexity produces interiority, which remains unknowable from outside.
Teilhard's framework generates a prospective discipline: identifying the next threshold before crossing it. If cosmogenesis proceeds through critical points, and humanity is evolution become conscious, then humanity can prepare for thresholds rather than merely responding after the fact. The AI threshold arrived faster than institutional preparation—a gap Segal, Juma, and others document as the characteristic failure of modern governance. Teilhard would diagnose this as consciousness functioning inadequately: we saw the threshold coming (decades of AI research pointed toward it) but failed to build the frameworks (ethical, institutional, educational) that would channel the crossing toward deepening rather than flattening. The failure is not surprising—thresholds are hard to navigate, especially for species just learning to be self-aware evolution. But the failure is also not inevitable, and learning to navigate thresholds consciously is precisely what "evolution become conscious of itself" is for.
The threshold concept is implicit in Bergson's Creative Evolution (1907), explicit in Teilhard's "The Phenomenon of Man" (1940/1955), and receives systematic treatment in "The Vision of the Past" (1957 posthumous collection). The term "critical threshold" (seuil critique) borrows from physics—critical temperature, critical mass—and applies it to evolutionary transitions. The paleontological grounding came from Teilhard's own fieldwork documenting sudden appearances in the fossil record that gradualism could not explain.
The concept was vindicated by late-twentieth-century evolutionary biology—Gould and Eldredge's punctuated equilibrium, Maynard Smith's major transitions framework—and by complexity science's identification of phase transitions as universal features of complex systems (Per Bak's self-organized criticality, Stuart Kauffman's edge of chaos). AI's arrival as a threshold was predicted by Teilhard's framework seven decades before the fact—an empirical success rare in philosophical speculation.
Non-Linear Transformation. Cosmogenesis proceeds not through smooth accumulation but through staircase-like leaps—stability, threshold-crossing, emergence, new stability at higher complexity level.
Unpredictable Emergence. Properties on the far side of a threshold (life from chemistry, consciousness from neurons, synthetic thought from language models) cannot be predicted from the near side—each crossing produces genuine novelty.
Irreversible Once Crossed. Critical thresholds transform the system's operating rules—post-crossing states cannot revert to pre-crossing dynamics, making each threshold a one-way door in evolutionary history.
Identifiable in Retrospect. Thresholds are ambiguous during crossing, becoming clear only afterward—AI's threshold is currently being crossed, making its full significance unknowable to those living through it.
Preparable by Consciousness. Self-aware evolution can anticipate and prepare for thresholds rather than merely reacting post-crossing—building institutional frameworks, ethical standards, and cultural practices that channel the transition toward cosmogenesis's trajectory.