Constructing the future of (digital) minds Home | Contact 
Research & Development

Traditionally, AI-systems are developed using one of two major paradigms: Symbolic AI on one side, and Artificial Neural Networks (ANN) on the other side. Both methods have severe practical barriers that have been attacked for decades now and neither have managed to overcome their main problems. Symbolic AI tries to develop algorithms for each and every problem that the perceived system must be able to handle in 'reality', and clearly fails because 'reality' is hard to predict. On the other side the ANN approach needs so much CPU-power, that any 'human level' results will be impossible for many decades to come. Besides that, ANNs are 'black box' systems; they are capable of 'learning' to recognize specific 'patterns', but we have no clue what a certain ANN actually has learned or recognizes, when it works.

The MIND|CONSTRUCT model borrows the strong points from both methods (generic approach, symbolic (semantic) knowledge modeling), while staying away from the obvious pitfalls of both methods (emulating low-level brain systems, trying to describe every functionality on its own). Our 'model' has both 'symbolic' alike properties and similarities to Neural Networks like 'weighted connections' and 'forward/backward propagation' (although the actual propagation is different from ANNs).

Building the AI-mind

Unlike the prevailing view in AI-research, being that intelligence has it's origins in complex processes, we are following the idea that intelligence is mainly based on a complex data-structure. To make a comparisson to our own brain; the human brain is not the equivalent of complex software, but is much more comparable to a very large interconnected database.

To achieve so called 'Artificial General Intelligence' (AGI), or even better 'self-conscious AI', it is necessary to build a machine that is a reflection of our own capacity to experience things, to understand and to reason. The traditional AI-views that are based on the idea of adding growing amounts of code (Symbolic AI), are clearly not the solution.

One of the most prominent 'hard problems' is the 'symbol grounding problem'. The hypothesis states that as long as you can only use symbols to describe other symbols, you can never reach real 'understanding'. To promote symbols to real understanding, it is necesary to 'ground' these symbols in reality. So what we actually need is a machine that is capable of 'experiencing reality'.

Development tooling

To be able to develop our product we need several tools, frameworks, libraries and other already existing technology. On this website we will document the step by step assembly of our development platform.

We use mainly Open Source tools and frameworks for our development. This means that the platform we are assembling will also be available directly for other researchers and possibly for students. By following this view we hope to stimulate those people that are interested in AI, or even want to make their career in AI-research and development.

Final product

The final product is a software-application that implements a 'consciously reasoning Artificial Intelligence' onto a wide variety of hardware platforms. This hardware can be a humanoid robot or a mars-rover, but it can also be a hospital-bed that can directly and appropriately react and act in specific situations.

The developed technology will be available, based on licensing, to be implemented into third party products.

Protection of Intellectual Property

The technological model of our AI-mind is specific, innovative and unique. Patent-applications (worldwide) are currently in preparation. Besides the patent-application, all involved organizations and people are held to a non disclosure agreement.

WhitePapers, research documents, etc.
Research papers

 • Why we need 'Conscious Artificial Intelligence' 2012-05-10
Project tooling
©2010-2015 MIND|CONSTRUCT - All rights reserved @Google+