Sebastian’s Predictive Machine

Physicist and independent researcher Jinesh K. Sebastian conceptualizes the brain as a predictive machine (Sebastian, 2025): a multi-layered feedback control system that generates future world-model states from learned priors by minimizing qualia potential and grounding predictions in sensory data. This framework draws on contemporary predictive processing theories (Friston, 2010; Clark, 2013), but extends them by incorporating qualia potential as a motivational driver, echoing the triadic gunas (sattva, rajas, tamas) of Sāṅkhya.

The architecture operates via a closed-loop hierarchy aligned with neural oscillatory bands, reflecting empirical findings on brain rhythms (Buzsáki, 2006). At the apex (delta center) of the loop, future world-model states (objects and actions) are generated from the current state, accumulated error, and learned priors under qualia-potential minimization. These predictions are sent downward as delta-band signals and decomposed into sensor-specific predictions (beta-band) at the beta center.

Predictions are sent to multiple gamma centers, each specialized for a sensory modality (e.g., vision, audition). Here, incoming sensory signals are compared with beta-level predictions, and modality-specific prediction errors are computed as gamma activity. Because each gamma center operates independently, this error computation occurs in parallel across modalities.

Error signals then propagate to the alpha centers, where modality-specific corrections are temporally integrated over a finite window, weighted to emphasize recent information. The length of this alpha-band integration window determines the persistence of perceptual representations, with the world model updated only after integration—giving rise to phenomena such as persistence of vision.

Because multiple alpha centers operate in parallel, their outputs (alpha-bands) are integrated at the theta center, which combines corrections over a longer temporal window spanning several alpha cycles. Modality-dependent weighting and thresholds enable coherent multisensory integration, accounting for phenomena such as the rubber hand illusion. This level also supports inference across missing modalities—for example, allowing a previously seen object to be represented when later inferred from sound alone.

The integrated correction from the theta center is sent back to the delta center via theta-band activity, where the world model is updated and new predictions are generated through qualia-potential minimization. In this way, top-down priors are continuously calibrated against bottom-up sensory signals through a delayed error-integration loop.

Within this framework, according to Sebastian, consciousness is analogous to the Seer (Puruṣa) of Sāṅkhya philosophy, with the key distinction that it is not eternal but emerges from neurological processes. It can observe the world model, experience variations in qualia potential, and embody itself within a self-model represented as an object in the world model.

From this perspective, all entities in the world model—including the self-model—are objects. When consciousness embodies the self-model, it becomes the locus of experience, with subjective perception arising from this coupling while other entities remain objects. If this coupling weakens or temporarily decouples, consciousness can experience the world model without embodiment, producing phenomena described as out-of-body experiences.

Oscillations in the predictive loop

In analog electronics, a feedback amplifier can enter self-sustaining oscillations when it meets the Barkhausen criteria of sufficient gain and appropriate phase. Sebastian argues that a similar instability can arise in the brain’s predictive loop, but within the Predictive Machine framework, structured perceptual phenomena result primarily from an expanded alpha-center temporal integration window rather than increased loop gain.

When the alpha integration window lengthens, prediction errors persist longer, and successive world-model updates overlap in time. This produces visual phenomena such as motion trails, stroboscopic effects, and frame-like persistence, explaining effects commonly reported after psychedelic consumption. These phenomena reflect extended persistence of vision rather than altered sensory input.

Related phenomena can also be observed in pathological cases. For example, individuals diagnosed with acquired savant syndrome, such as Jason Padgett, report perceiving the world in discrete frames and visualizing geometric structures. Within this framework, such experiences can be interpreted as consequences of altered temporal integration and predictive updating in the perceptual loop.

Because the predictive loop constructs a coherent world model, resonant dynamics within it can organize perception into stable geometric modes. Constrained by boundary conditions and hierarchical structure, these modes produce patterns analogous to Chladni figures, Faraday waves, and cymatic resonances, offering a unified explanation for the luminous geometric forms, mandalas, and chakra-like patterns reported in deep meditative and psychedelic states.

Qualia potential

Analogous to surprise (variational free energy) minimization in active inference, Sebastian’s theory introduces a scalar qualia potential. Surprise minimization proceeds in two stages: first, world-model mismatches are resolved through sensory grounding; second, qualia potential is optimized to guide future predictions and action selection. This separation addresses the dark-room paradox (Friston, 2012) and aligns with observed human behavior, including motivational shifts, sustained goal pursuit, and homeostatic drives (Solms, 2021). Notably, a concept analogous to qualia potential appears in Sāṅkhya philosophy as the gunas.

Sebastian states that the Predictive Machine Theory accounts for a range of subjective phenomena, including persistence of vision, meditative and psychedelic experiences, and multisensory illusions such as the rubber hand illusion, while establishing a conceptual bridge between ancient Indian philosophy—particularly Sāṅkhya—and contemporary frameworks such as predictive coding, the Free Energy Principle, active inference, Global Workspace Theory (Baars, 1988), and Solms’s neuropsychoanalytic theories.

Thus, the Predictive Machine models consciousness as an emergent process observing the world model, experiences variations in qualia potential, and can identify with a self-model embedded within it, offering a formal account of subjective phenomena emphasized in Eastern philosophy.

References

Sebastian, 2025 Jinesh K. Sebastian, The Predictive Machine - How Minds and Robots Build Reality from Within. Zenodo - ttps://doi.org/10.5281/zenodo.17387527

Baars, 1988. Bernard Baars, A cognitive theory of consciousness. Cambridge University Press.

Buzsáki, 2006. György Buzsáki, Rhythms of the brain. Oxford University Press.

Clark, 2013. Andy Clark, Whatever next? Predictive brains, situated agents, and the future of cognitive science. Behavioral and Brain Sciences, 36(3), 181–204.

Friston, 2010. Karl Friston, The free-energy principle: A unified brain theory? Nature Reviews Neuroscience, 11(2), 127–138.

Friston, 2012 Karl Friston, Christopher Thornton& Andy Clark, Free-energy minimization and the dark-room problem. Frontiers in Psychology, 3, 2012.

Solms, 2021 Mark Solms, The hidden spring: A journey to the source of consciousness. W. W. Norton & Company.