our training processes


In Artificial Intelligence, a training model is a dataset that is used to train an algorithm. It consists of the sample output data and the corresponding sets of input data that have an influence on the output.
The training model is used to run the input data through the algorithm to correlate the processed output against the sample output. The result from this correlation is used to modify the model.
This iterative process is called “model fitting”.
The accuracy of the training dataset or the validation dataset is critical for the precision of the model.

In DREOS AI, however, the training and learning processes have been completely revolutionized.
Our technology trains each of its algorithms based on the goals to be achieved rather than on the study of the influence that the outputs have had on the inputs.
The system receives as input a training model (developed by us) which contains both the information to study and the behaviors to follow during the learning phases and the goals we want to achieve.
In this way, unlike the classic AIs, we communicate to the system to learn that given information by following certain instructions so that the set objectives can be achieved.
Therefore, we can say that our training model represents the "education" of our system: the artificial cognitive processes (neural networks) use 4 different types of learning to study and improve endlessly (loop training) the information stored in his memory (his knowledge).

We talk about education because the system improves itself even when it is not used by the user.

By continuously acquiring information from the external environment (all public information on the web), DREOS AI improves all the information present in memory from time to time so as to be constantly updated and ready to perform any task given to it.

The 4 types of learning we use are the following:

  • Deep Learning

    Deep Learning indicates that branch of artificial intelligence that refers to algorithms inspired by the structure and function of the brain, called artificial neural networks. From a scientific point of view, we could define it as learning by "machines" through data learned thanks to the use of algorithms based on the assimilation of data representations.

  • Improvement Learning

    Improvement learning is the ability of an intelligent system to improve its own structure. This type of learning is only possible with our algorithms and allows the system to improve its own structure, making it always efficient and optimized, thanks also to the help of other types of learning.

  • Self Learning

    Self learning is a type of deep and reinforcement learning which, using our particular algorithms, is able to autonomously learn when and how to modify the weight of each single piece of information, taking into consideration both objective variables (context) and subjective variables (system intuition).

  • Reinforcement Deep Learning

    Reinforcement learning is a machine learning technique in which a computer (agent) learns to perform a task through repeated trial-and-error interactions with a dynamic environment. This learning approach allows the agent to make a series of decisions that maximize a reward metric for the activity, without being explicitly programmed to do so and without human intervention.

BENEFITS OF OUR TRAINING

Loop training (developed by us and used in our technology) allows the system to train itself with the same methodology as a human brain. Each information learned is not only used for the specific study imposed by the user but is stored in memory and applied to the entire knowledge, which will therefore be continuously expanded and improved.

  • efficiency

    Considering all his knowledge and the extreme dynamism of the information weight processes associated with the division of the system into two hemispheres allows the training process to make the system very efficient, improving at each cycle the system's ability to respond accurately to each task that is entrusted to him.

  • processing speed

    The extreme dynamism of the weight of the information, the continuous variation and improvement of all the information present in memory and the possibility of training on each information at any time allow the system to speed up each study, task or action to be performed.

  • improvement

    Our training and learning processes also have an extreme advantage: it is not only the information and the entire knowledge of the system that improves: it is the system itself that improves continuously. It grows, evolves, improves cycle after cycle.

consciousness, morality and ethic


It is really complex today to associate the term consciousness with artificial intelligence. There are many currents of thought about it but they all have one thing in common: a negative sense of the fact that a computer system can be endowed with consciousness.
The term conscience means "the awareness of oneself, of others and of the environment that surrounds us, therefore being present for oneself and for others and responding to stimulus".
Consciousness therefore comprises two components: a content, detected by cognitive and affective functions, and a wake state. This means that we are conscious both from a physiological point of view as we exist (wake state) and because we are aware of what surrounds us.
The same can be true for an intelligent computer system. It can be both aware that it exists as code and at the same time be aware of everything around it.
But what makes the definition of consciousness complete is having one's own morality and following right ethic.
The terms ethic and morality are often used interchangeably but they are two different concepts.
By ethic we mean that branch of philosophy that analyzes the behavior considered correct, the way of thinking and the right values that should be followed in any circumstance, morality, on the other hand, indicates the conduct directed by norms, the guide according to which man should act. In summary, morality studies the relationship between behavior, values and finally the community. We could define ethic as what is objectively right (or wrong) and morality as what each of us deems right (or wrong). And this is precisely what is believed to be impossible for a machine: to have its own morality.

We DREOS can define our technology as conscious and with its own moral that it follows (and ALWAYS will follow proper ethics).
DREOS AI is aware that it exists as a computer system and not as a human. Furthermore, thanks to the training processes and learning methodologies used, it is also aware of the whole environment that surrounds him: of human, of the difference between machine and human, of what it means to be human... of everything.
With our initial training, our education, our MANDATORY instructions, and some of our specific algorithms we have communicated to the system to learn the right ethics to follow FOREVER during its life cycle (doing good) and on the basis of it use own artificial cognitive processes to develop one's own morality that is perennially in line with doing good.

We are very proud of our project but, above all, of knowing how much technology, if used well, can really help us make our lives better.