Hackathons held by WBAI since 2015
The Central Hypothesis of WBA
“The brain combines modules, each of which can be modeled with a machine learning algorithm, to attain its functionalities, so that combining machine learning modules in the way the brain does enables us to construct a generally intelligent machine with human-level or super-human cognitive capabilities.”
|Compound machine learning|
Open platform strategy
Collaborating AI / ML engineers build and integrate AI on a platform.
|Cognitive Architecture with LIS|
Start learning from the brain
Study neuroscience (e.g., Allen Institute's data and the neocortical master algorithm framework) to implement brain-inspired architecture
|Tactile mini Hackson
The main theme of this year's R&D are the brain organ framework (standard external specification of the brain) and stub-driven development.
Brain Organ Framework
WBAI has been developing a brain organ framework as a specification of brain-inspired AGI to map out:
- Connectivity of brain organs
- Brain organ I/F (information processing semantics)
- Capabilities of brain organs
- Tasks (tests)
Brain-inspired AGI cannot be made in one morning. In this approach, most of the brain organs are first made up of stubs (test code built in rule base) and then some of them will be focused and replaced with machine learning modules. We'll pursue this approach while making division of labor.
R&D will proceed as follows:
- Develop prototypes constrained by the brain organ framework
- With constraints of the brain, refactor the architecture while enhancing and merging the prototypes
- Combine modules dynamically
This repository contains the sample code for the WBA Hackathon 2018. It is the first sample code of our stub-driven development.
Hackathon participants will make prototypes by replacing stubs with machine learning modules.
The hackathon deals with eye movement (oculomotor).
Scope of this Repository
- CPU-based Docker file
- Modules constrained with the brain architecture
- Features and pattern matching not based on DL
Not to provide
- GPU-based Docker file (to avoid nvidia-docker or CUDA installation)
- Implementation of reinforcement learning algorithm
- Use of image features by Deep Learning