in Artificial Intelligence, Freudian Artificial Intelligence

“It was scary stuff, but radically advanced. I mean, it was smashed, it didn’t work, but… it gave us ideas, took us in new directions. I mean, things we would have never… All my work was based on it”

Terminator 2

Does artificial intelligence coincide with neural networks?
Certainly not, even if the advertising often wants to make the equivalence appear to be true.

In our research and development work on AI issues, we are investigating paths alternative to the most commonly used methods.

The idea is to render (at least in part) algorithmically a theory of the mind with a long history.
The theory we are working on was called Metapsychology by its creator, Sigmund Freud, in the last century.

In his works, Freud transforms the evidence on human behavior into a dynamic theory (exposing and evolving it in thousands of fascinating pages) that refers to classical physical theories, both in terms and in content.

Is Freud’s theory to be considered a correct model?
The answer is difficult to give as the Freudian model is not formalized. The model is certainly plausible and captures different analogies between mental dynamics and physical models that make it attractive.

In our attempt to formalize this model in mathematical terms, we have managed to grasp some components that have already allowed us to enrich and make our algorithms more effective. These innovations have proved to be effective on different types of data (mainly text and images) without great need for adaptation.

Regardless of the ability to predict the behavior of the mind, his inputs have therefore already had positive repercussions in our work that convince us of the validity of this idea.

The road is long but, given the articulated structure of metapsychology, we think that it is possible to build, along the way, different algorithms with innovative features compared to the existing ones.

For us, in addition to the intellectual appeal of this strange challenge, there is also the prospect of having tools more understandable in their operation than neural networks (critical point of the latter) with a lower computational cost and easier to extend and adapt.

Do you want to know more and collaborate or support this journey? Contact us