[Article.Ai] What is next after deep learning?

Deep Learning has evolved, is still evolving, and will likely continue to evolve.

Deep Learning may go in the direction of Supermathematics, or Euclidean Superspace, because there is:

(1) Evidence that the brain does some form of Supersymmetric operation.

(2) The reality that Deep Learning uses cognitive science rules in the neighborhood of (1) to bound models in some way, like how Deepmind limits some of their models in terms of certain cognitive science rules.

▽

▽

Supermathematics and an experimental hypothesis:

For example, based on the Quantum Boltzmann Machine, and Quantum Reinforcement Learning, (See Video for Quantum Boltzmann Machine) I organized a simple hypothesis for implementing Deep Learning in Euclidean Superspace, on a quantum computer (i.e. Dwave system).

See a clear overview of the hypothesis: Supermathematics and Artificial General Intelligence

See the hypothesis: Thought Curvature: An underivative hypothesis

See a discussion regarding the hypothesis, on a science board here.

▼

▼

Notably, the hypothesis entails a likely plausible way to run Deep Learning, in the regime of Euclidean Superspace, by generalizing from the Transverse Field Ising Spin Hamiltonian operation seen in the video for Quantum Boltzmann Machine, to some form of (Super-) Hamiltonian sequence:

Footnote — limitations of the aforesaid experimental hypothesis:

Although the hypothesis may turn out to be invalid in its simple description, there is a non trivial possibility that the math of Supermanifolds may inspire future Deep Learning. Recall that cutting edge Deep learning work tends to consider boundaries in the biological brain, and biological brains can be evaluated using supersymmetric operations.

▼

▼

Author:

I am a casual body builder, and software engineer.