“General Purpose Learning Machines was thought to be theoretical impossible until now…”
Featuring ISRA protocol, an experimental Artificial General Intelligence architecture
Design
The system is capable of representing any kind of logic, reason, experiences, sensory input, basically any possible combination of thought one’s mind can generate in a simple unified generic memory structure.
It learns by unsupervised trail and errors in real setting, gaining experiences through imitation and repetition, and improve decisions accordingly.
Overview
Quick overview of all the main operations and how they relates to each other (12sec):
Video doesn’t cover:
- How decision is made in depth.
- How the system make predictions, learn from failure and maintain learning potentials (fitness function).
- Algorithm used for pattern matching.
- Internal experience based memory structure generation and management.
Participate
The project prototype already exist in github, but due to current lack of development fund and aids, the code is not optimized and only support a few hardcoded hardware.
Therefore real time demo footage is not yet available.