Percipion : making machines smarter using cognitive computing

All individuals exist in a continually changing world of experience (their phenomenal field) of which they are the center. This perceptual field is a reality for the individual. Through multiple dedicated neurophysiological hubs, what individuals perceive will be decoded and transformed to be acquired and further processed or will be forgotten. Here, the long-term memory is a key hub, located deep within the white matter of the cerebellum, since it will store knowledge, better said our knowledge so that the best vantage point for understanding behavior is from this “internal frame of reference”.

While they communicate, individuals will freely associate specific words together and describe, using their language, using their knowledge, how they perceive, represent, or understand a given situation.

As a human, our ability to reason, make decisions, and solve problems can define intelligence. These cognitive processes are also linked to psychological functions such as emotions, sentiments, or needs. These mental states provide additional context-related vital insights, which, once added to other available contextual inputs, will create a conglomerate of information, giving us a mental representation of the situation (understanding), enabling our awareness, and adapting our behaviors and actions.

Making machines smarter will require that these human capabilities can be transferred to and executed by a machine.

Since 2016, our research and development team, is working on this topic to create a Cognitive Computing system that is providing machines with a mental model of the surrounding world to be smarter, and able to replicate the way humans understand.

This system has to be able to process information like it is processed in our brain by our chemical neurons, in an analog way or fuzzy way so that this processing is as close as possible to how humans process information (Aeteos SmartNeuron? software component). In addition, it is also key to provide machines with a data structure analog to the human long-term memory, where information or knowledge is stored in a categorized way, providing machines with specific frames of references and enabling cognition (Aeteos VirtualBrain? software component).

Empowering machines to understand is the first cognitive tasks that needs to be implemented. Natural language understanding technologies involve identi?cation the intended semantics from the multiple possible semantics. That means the system needs a contextual lexicon of the language, with a suitable ontology, to provide machines with a mental model of the surrounding world and assess the provided data to extract its meaning in several languages without prior training (no machine learning). This contextual ontological lexicon is then added to a categorized symbolic object structure so that machines can better understand, be smarter, and able to replicate the way humans understand.

At Aeteos, this is what we do and Percipion is our Cognitive Computing software solution. We don’t think that any statistical approach will make machine smarter. The human brain doesn’t work like this, the way humans learn is far more complex. The only way forward is to decode stimuli from our phenomenal field in such a way that it can be decomposed and linked to a given prototypical symbolic knowledge object. Cognition is only possible under these conditions.