Revealing the secrets of intelligence

Last week the Springer SN Computer Sciences published my last paper related to the research of intelligence. It is a deep dive into the learning process of the neo-cortext. The algorithm uses many findings in neuro sciences and does not just make use of heavy use of mathematics as conventional "AI" algorithms do.

In this at least to me, highly interesting research in the field of AI, it becomes clear why a brain has to go through a baby phase and what happens when this is not the case. How does a baby learn in the first place? A learning algorithm for learning a pattern, namely the Spatial Pooler, shows how the neocortex presumably stores and recognises diverse patterns. Spatial pooling is an algorithm that learns spatial patterns.

This newly explored algorithm is modelled on the way the brain works. The conventional algorithms do not appear to be intelligent. They only solve the problems they are designed for. very successfully. AI is often programmed to do a specific activity. In order for the AI to learn these activities, thousands of images of specific actions must first be read in, which takes a lot of time and effort.

This is precisely the issue that distinguishes the ML approach from AI. A conventional algorithm that recognises images, for example, cannot recognise music or learn sequences, etc. A brain has a universal algorithm (cortical algorithm) that can be trained for all kinds of tasks.

I have been intensively studying the functioning of the brain for the last few years because I am convinced that we have not yet found the true AI. For far too long, we have been dealing with purely mathematical models and statistics. Such approaches work in practice in certain scenarios. But that does not make such algorithms intelligent. In my opinion, the secret of intelligence lies in nature itself. I learn from biology and neuroscientific findings and work on algorithms that artificially reproduce the biology of the neocortex as closely as possible.

Based on ideas from Hierarchical Temporal Memory, I have developed a framework in .NET Core called NeoCortexAPI. This open-source project contains all the research results in the form of a C#/.NET API.

If you are interested in this work, feel free to read the paper. I’m happy to answer all your questions. I know, it is a hard read, but the topic is more than just high-tech science. It digs deep into the understanding of our observation of the world around us. To me, the question about understanding intelligence is also a philosophical one. But, to be truly able to understand how intelligence works, I believe it is important to dive into the neuronal level to be able to understand how single neurons are organized in a population that can build cognitive capabilities.