Artificial Intelligence Colloquium: Physics of Artificial Intelligence

Speaker: Mr. Ted Senator, Program Manager, DARPA / Defense Sciences Office

DARPA is exploring how to incorporate physics knowledge about complex dynamical systems into AI learners, especially deep networks to overcome some limits of current AI. The physics knowledge may take the form of differential equations, invariants, or symmetries and may be incorporated into the structure of a deep network or into the optimizations used in training. This talk will describe several approaches and applications in structural dynamics, turbulence, biochemistry, biophysics, and complex natural systems.

[aigen-journey]

I know this is a stretch, but I can't find the research mentioned at 12:10
"unsupervised pre-training via prediction model using synthetic data" doesn't seem to be the title.