We Understand!
Syntience Inc is an independent and
privately funded Machine Learning
and
Artificial Intelligence Research company
founded in 2004.
We are focusing on
Deep Neural Network based
Natural Language
Understanding (NLU).
We have developed a superfast algorithm for learning human languages called Organic Learning (OL).
Under certain constraints we can claim that
OL can learn a useful amount of any language
on the planet
in five minutes of unsupervised
learning from a plaintext 5MB Unicode
corpus when running on
an old x86 laptop
without using a GPU.
We are productizing this capability as a RESTful cloud server named Understanding Machine One
Below we discuss our offering and some end
user oriented applications we could implement
in order to demonstrate UM1 capabilities
The high level results of our 20+ years of research are discussed in our blog and on our video page
Our language learning algorithm is described in Chapter 8.
It uses a Deep Discrete Neuron Network where
discrete Java Node objects are used as
Epistemologically Adequate pseudo-neurons.
The system starts out empty and adds "Neurons" as
required when learning from the given corpus. This
is very different from Deep Learning.
UM1, our Understander (our runtime inference engine) is described in Chapter 9 of the blog and is available as a free alpha service in the cloud. You can test the service yourself by downloading our python test code from GitHub
The long term goal of the company is to create a portable personal AI which uses voice for input and output, as seen in the movie "Her" and as discussed in Science Fiction for decades.
In the short term we intend to use our UM1 Understander as the basis for a novel message routing system called Bubble City. It is not a social medium, it is an Idea Router. The current design proposal is available as a PDF.