We develop the GAIuS Cognitive Computing framework for evolving Machine Intelligence/AGI

Learn more about our Technology

Cognitive systems are designed to mimic human brain processes

Like humans, cognitive systems learn as they go along. They learn from results of actions taken, and apply that knowledge when a similar situation emerges. 

Unlike humans, a computer is able to use almost unlimited information and its decision-making is therefore based on a wider range of evidence.

A key feature of cognitive systems is that they come to understand their environment, and adapt as information, goals and requirements evolve.

Cognitive computing relies on machine intelligence, but with additional processes that allows it to anticipate and solve more challenging problems. These integrated features give cognitive systems the ability to cope with uncertainty that would cause problems for conventional AI systems.

GAIuS is the next-generation machine intelligence engine that shifts human programmed to automated machine-evolved solutions

GAIuS is the convergence of different technologies into a coherent solution to machine intelligence. It provides the only viable pathway to truely intelligent, autonomous, evolving machines. This yields unique opportunities for data-driven solutions. 

agi_features.jpg

The Convergence of Artificial General Intelligence, Deep Learning and Big Data

 

General Artificial Intelligence

The most important, fundamental component of GAIuS is its Cognitive Processor.

The Cognitive Processor learns from incoming data, extracts information, recognizes what it’s observing, compares and contrasts it with what it knows, predicts what will happen, and makes decisions aligned with its goals.

The Cognitive Processor does this regardless of the problem domain. Meaning, this process is effective whether the application is intrusion detection, stock market predictions, or machine vision.

Part of GAIuS’s power is that it doesn’t model the data at training time. Modeling data at training time prevents real-time learning and causes other issues like "overfitting". Instead, GAIuS indexes data and allows modeling at prediction time. GAIuS continues learning from real-world data even after the training period. As an added benefit, it won’t suffer from Catastrophic Forgetting, which is a major limitation of other AI systems.

Cognitive Processors consist of one type of node in a “network topology”. Other nodes provide flexible operations to incoming or outgoing data. Different connections between different nodes change the way information is processed by the GAIuS Agent.

 
artificial_general_intelligence.gif
 

 

Deep Learning

When these nodes are wired in a hierarchical topology, they abstract information from noisy lower layers providing higher layers with more consistent values. This is called “deep learning”. GAIuS does this without suffering the problems of artificial neural networks (ANNs) or convolutional neural networks (CNN). GAIuS doesn’t require that you know the number of classifications in your data or environment ahead of time. In fact, your environment can change by needing more classifications. The GAIuS Agent will automatically adapt and learn these new classifications!

 

Big Data

GAIuS flourishes on Big Data. The more real-world data your GAIuS agent observes, the better it becomes at its job. GAIuS is an “information processing” engine. It takes raw data, and extracts useful or actionable information from it to discover patterns and make predictions.

Does your data contain noise? Don’t worry. GAIuS can figure out how to separate the signal from that noise. It can even suggest data filtering options to improve the signal-to-noise ratio, or “SNR”.

Don’t know if your data contains any useful or actionable information? Again, GAIuS has the solution. It’s modern information processing algorithms will quantify the amount of useful or actionable information! It will also show you the profile of the kind of data that you have, whether it is robust, trivial, or random as well as the key indicators in your data.

GAIuS agents are created by making a “network topology”. With standardized inputs and outputs, your GAIuS Agent can accept any type of data that your application will encounter in the real-world, including vectors, strings and data from other GAIuS agents! Agents can be networked together to share and process information.

GAIuS unique architecture allows modularity and separation of cognitive processing and data processing components. Any solution can be derived from these components, or easily integrated with your own pre-existing solutions in the application layer. Here are some things GAIuS can do:

 

RECOGNITION & CLASSIFICATIONS

A GAIuS agent can recognize what its seeing. It can be used to “classify” objects based on data.

 

PREDICTIONS

GAIuS agents can make predictions of future events and their utility (i.e. how good or bad that future is expected to be).

 

ANOMALIES: MISSING AND EXTRA

GAIuS provides predictions about the future, as well as, the un-observed past. It describes the “present” state (i.e. recognition & classifications), and returns anomalies, i.e what is missing &/or extra in the current observation compared with what it was expecting.

 

DECISIONS & ACTIONS

Three factors effect your GAIuS Agent’s behavior:

  1. The genome.

    The genome is your GAIuS agent configuration. It consists of various connections and parameters that can be changed for the specific environment or problem the agent faces. Changing the genome changes the agent’s behavior.

  2. Your data.

    Whether your data is streaming live, provided manually, or given as a bulk training session, your GAIuS agent will learn from it. Unlike other techniques, learning isn’t a one-time event for your agent. It will continue learning new data, in real-time. No a priori classifications are needed. No knowledge engineering is needed. And you don’t need to know anything about your data before giving it to an agent. The agent’s behavior adapts to the data it receives.

  3. Feedback from you or its environment.

    Your GAIuS agent can be trained to change its behavior through operant conditioning or reinforcement learning. The agent changes its behavior to achieve the goals you or the environment have provided through positive or negative feedback.

The GAIuS framework creates an agent that adapts to its world in real-time. The agent’s modularity allows simple elements to be wired together in different ways to produce complex behavior.

Our Solutions Evolve with Genetic Algorithms

 

Your GAIuS AGENT can evolve

It is impossible to manually program intelligence that can adapt to any situation. There are far too many variables and situations to consider. No army of engineers nor programmers will ever be able to write code to handle the multitude of possibilities.

Nature has solved this problem by creating elements that evolve based on their environment.

So, too, GAIuS is software that evolves. We use Genetic Algorithms to mutate their parameters to automatically create new, novel solutions.

You can also breed solutions together - multiple GAIuS Agent's can be created and the best be bred together to produce offspring with desirable characteristics. This process is similar to breeding of animals to produce behaviors that humans find useful.