seminars

Natural self-assembly and spontaneous dynamics in brains

Date
Wednesday, April 3, 2019

Time
12.00 p.m.

Location
ISI seminar room 1st floor

Speaker(s)
Dr. Jacob Billings - Emory University, GA (EU)

From one perspective, time and space are generative processes enumerated by state changes among spontaneously fluctuating and interacting quanta. Indeed, the most probable next state of a system is a direct function of its previous interactions. The next state’s timing and location also may be written as a function of previous interactions. For quanta whose interactions once achieved a particularly low-entropy state, “an arrow of time” points to increased universal disorder with each new interaction. The presence of local thermodynamic gradients may, however, spur the entrainment of adjacent quanta into patterned structures that minimize the local accumulation of entropy.
As cells evolved, ribonucleotides helped minimize the loss of useful chemical information. As multicellular animals developed brains, their sensory processes better coordinated the maintenance of homeostatic drives. Relative to some other mammals, homonids developed an expanded 6-layer cortical surface in a patch of the dorsal telencephalon. Interestingly, neo-cortical expansion in homonids left the primary sensory areas to approximately an equal distance from one another. This structured expansion may maximize multisensory processing.
Wheres speciation through many generations produces gross changes in brain morphology, all brains are varieties of highly specialized gels; hard-wired for particular computations and yet plastic enough to negotiate rapid changes. Advances in whole-brain imaging has allowed unprecedented access into the spontaneous structure and function of the space occupied by brains. By taking note of recurrent patterns in brain hemodynamics, researchers have begun to decipher intrinsic flows of information among brain nuclei. Brains appear to embody a small-world architecture that maximizes the ability to synchronize large numbers of neural cell bodies with minimum message conduction costs. The resting state of the brain is one of habitual activity. And, regardless of the kind of external stimuli, the brain expresses large-scale spatio-temporal activity cycles every few moments.
The presence of structured rhythms in brains, alongside the brain’s ready capacity to adopt novel states, underscores the hypothesis that spontaneous self-assembly in biological systems is an intrinsically dynamical accumulation of information-rich, low-entropy, events. Looking forward, the increased capacity of modern deep-learning paradigms to compute arbitrary functions via observation makes the task of modeling brains in silicon to be almost an engineering challenge whose goal is to assemble some appropriate collection of co-dependent neural nets that, together, approximate observed brain dynamics.