Using the Information and Communications Technology Data Deluge from a Semantic Perspective of a Dynamic Challenge: What to Learn and What to Ignore? -Part 3-

The paper analyses the Data Deluge premises and main approaches toward a dynamic and semantic search of balance between generating data and extracting knowledge, focusing on some semantic criteria of the search, using Information and Communications Technologies (ICT) context at the planetary scale, which could influence Earth environment and eventually humankind evolution with an unprecedented speed and impact. The main reasons for this kind of analyses come from the ICT generated global and exponential changes, which are more and more difficult to be controlled or at least estimated, regarding all their positive or negative consequences, just because of the complexity and speed of their vast penetration. The semantic analysis have to face a dynamic environment, because the relevant data/facts are fast changing (not from centuries or decades, but from months or even days), so we have to refine our search looking for new semantic terms or changing the old meaning (associated information/ knowledge) of the usual ones. It is emphasized that, although all technologies and ICT applications largely contribute to DD, lately, AI and its main branches, machine learning (ML), deep learning (DL) or cognitive learning (CL), are prominent in striving to extract information/knowledge from data, or even to compete with the human intelligence by cognitive features. The work identified criteria of the search, which are linked with all (positive or negative) consequences of ICT on Information society (IS) toward Knowledge Based Society (KBS) and generally of World changes, just because of the complexity and speed of their vast proliferation, more visible by the dramatic climate changes, Earth resources fading, social imbalance or actually by the vulnerability of humankind to diseases spreading in the globalization context. Such consequences should be alarm signals and optimization criteria in our decision processes, as individuals, but mainly for the incumbents. These signals and the ICT impact on IS/KBS at planetary scale could indicate the main criteria for the mentioned semantic search in the ICT/AI/ML/DL/CL fast implementation processes, in order to meet Earth ecosystem needs and expectations for a sustainable progress. An important paper conclusion is that the semantic approach of decision processes, based on the generated data and extracted information, is a complex optimization target which just could match and benefit from the dynamic ICT/DD/ML/DL/CL context, where the data signification, implications and consequences are fast appearing and changing, because of this context potential to produce changes, but we have to be sure which of them are positive and which negative, to what extent and time period. The most difficult and complex problem will be, every time, to understand and define criteria for what is positive and what negative, as all our Earth ecosystem and even humankind behaviour and evolution is changing, but also our criteria should! Such conclusions are sustained by some concrete examples of the massive expansion of ICT, like the case emergent 5G (among others), which could produce, besides positive, real and unprecedented benefits, new, but less desired effects. Here, the good news is that AI/ICT itself could compensate such negative effects, as it is revealed also by some concrete applications cases. The second section of the paper completes the previous technical elements with other improvements of learning from the real-time complex processes of ICT/IS/KBS that could be observed just in the most prominent processes of learning at planetary scale, by ML/DL. Although we speak about how the machines are learning, these could reveal not only the state-of-the-art of ICT/AI/ML/DL/CL and their limits (for example, ML and Wolpert famous no free lunch theorem), but indirectly shows the limits of humans that have the responsibility to design them, just considering the mentioned criteria. Another conclusion is that all we see or perceive on these ICT/DD evolutions is not simply to understand nor to manage, but we have to timely analyse and try to optimize, as the price is so big. In fact, the complexity obliges us to also change the rules of learning we are used to practice, because the old rules could not be completely useful in the deep and ever-changing context, i.e., optimally and dynamically refine the criteria and the models of our thinking, in order to rapidly adapt to the new realities, which exceed the ICT/DD context, as we can see in these days, when the main challenges of our Earth ecosystem seem to come from other side (human and social health), but we also have to agree that any dramatic evolution at Earth scale have to be seen in the globalization context and we must use the power of ICT/DD/AI to optimally face the challenges (including the emotional impact on human life and behaviour, which ultimately could also induce, on short or long time, important changes, that we not really or completely perceive now). The final conclusion is that, for learning to better learn, the emphasis must be secondary on the fact that the human contribution to ICT/AI/ML/DL/CL processes is needed for better results against the difficulty/complexity, but primary in order to be sure that the desired outputs are designed to match the priority criteria we have above mentioned as essential for a sustainable development of ICT/IS/KBS and humankind evolution on Earth. More than these, learning to better learn, for a systemic approach of refining knowledge, is necessary to be the goal of all responsible people, beginning with the incumbents, the specialists and teachers, as everybody has to adapt to changes (Intelligence is the ability to adapt to change - Stephen Hawking) and use new technologies with appropriate criteria of optimization, i.e. timely analyse what and how we should learn or ignore.