In 2008 I decided to start a research and software company, Brainpower Labs. The main purpose of the company is to develop advanced AI and machine learning models and algorithms. A secondary purpose is to develop software to help fund this research.
The company started with an ambitious vision: to develop "advanced, human-like artificial intelligence software based on the architecture of the mammalian brain." Essentially, my initial plan was to single-handedly solve AI, explain the brain, and build a company around the results. This turned out to be difficult. So I redefined the company roadmap a few years later. Now the near-term focus is fundamental machine learning research, especially deriving new foolproof learning algorithms from simple, elegant principles. The long-term goal remains essentially unchanged: to build general AI.
So far, most of the effort here has focused on basic research and internal software tools. Some projects funded by the company include:
Binary Log-Linear Models: basic research on a powerful class of machine learning models, information theoretic priors, and a range of practical learning algorithms.
Anvil: C utility library used by all other Brainpower Labs projects; provides a wide range of basic utility functions, including machine learning, data processing, 2D/3D visualization, linear algebra, image processing, profiling, parallelization, window management, GUI, audio, file compression, and more.
Sapience: brain-inspired AGI architecture design, C++ and Python implementation, real-time probe for assessing performance, and a range of simulated test environments in 2D and 3D.
Compression AI: AGI research based on general purpose lossless data compression.
Sample Lab: interactive tool for comparing Monte Carlo sampling algorithms for discrete graph-based statistical models, especially focused on exact sampling algorithms.
SOM Lab: interactive visualization for topographic map learning algorithms, including various 1D/2D/3D lattices, test data distributions, and performance metrics.
On the funding side, see iBonsai, a meditative interactive 3D tree simulation, released for iOS.