Artificial Intelligence (AI) is the order of the day, present in many headlines about scientific innovations and novel applications in all types of phone number list . Even so, the AI concept sounds complex and very remote to us, few know that we actually live with it in the most everyday actions.AI is defined as the ability of machines, processors and software to mimic through algorithms functions traditionally associated with the human mind , such as perceiving, reasoning, learning or making decisions.In this definition lies the main difference of AI with a traditional computer program, since it acts according to precise instructions (if A happens, execute X, if B happens, execute Y), that is, there are orders that cover all the casuistry to which can be faced. The AI, on the contrary, performs actions without cause-and-effect orders based on previously entered data, which will be the foundation of its reasoning.Therefore, the results that this technology reaches will depend on the data initially entered (its quantity and quality) and the training to which it is subjected to strengthen its learning. Once assimilated, the AI-powered software in question will be able to draw conclusions on its own, and even continue to learn as it encounters new situations. This is what is known as machine learning .
AI in our day to day
Nowadays it is common to have smart speakers at home capable of being controlled by voice, find chatbots on web pages that answer any questions, or even unlock our Betting Email List using the fingerprint or simply the face – that was so comfortable before they appeared. the masks.Didn’t I mention that AI was more present in our lives than we thought? Voilà. Behind each of these processes there is neither a human being nor an infinite series of precise orders, but an algorithm that analyzes the data provided and reaches one or another conclusion . Depending on the case, this will be a response to an oral or written request, or the unlocking or not of the phone depending on the fingerprint or face analyzed.This technology is also present in areas such as industry, medicine or robotics, and in processes such as autonomous driving or decision support systems . The latter are based on the collection and analysis of large amounts of data and, based on this, they evaluate situations and probabilities and recommend how to proceed. Currently they are already being used around the world to support decisions such as granting or not granting a loan, or granting parole to a prisoner.So, recapping: there is a technology called AI that receives training based on data and supervised trials and, after these, analyzes situations and arrives at solutions autonomously. These conclusions largely intervene in decisions, some as decisive as whether or not to grant a person freedom, based on past data and probabilities. Let’s stop here.
Ethical implications of AI
woman with black face holding white masksAs we have discussed previously, the conclusions reached by an AI depend on the amount of data we provide it, but also to a large extent on its quality. What if programming facial recognition software entered a large amount of data, but it was not diverse enough? The answer is simple: this one would work much better identifying faces that you can compare with what was entered in the beginning, and it might not even work with those of which you have less information.This is what happened to Joy Buolamwini , a computer scientist and MIT researcher: She found that Amazon’s facial recognition software did not detect any faces on her , but that it did if she put on a white mask. With this finding, Buolamwini began to pull the thread and discovered that the system recognized the faces of Caucasian men perfectly, but that it was not able to identify women of the same race as well and it worked even worse with individuals of non-racial races. white.Despite the fact that when an AI begins to work it is perfected through machine learning , the initial data on which its algorithm is built is crucial and inescapably determines its performance, whatever time passes. For this reason, Buolamwini decided to investigate the implications that building AI algorithms on skewed data could have, as explained in the documentary ” Coded Bias ” that can be seen today on Netflix.It also raises another aspect to consider, much more important than the bias in face recognition: what about decision support systems? They are based on past and current data regarding socioeconomic characteristics, crime rates, and all related to sociodemographic data such as age, gender and race. What if the data on which we base them have a racial bias , caused by the socioeconomic inequalities that exist between races? What if these socioeconomic inequalities disappeared in the future, but we continue to use the same decision support systems? Would they have become hopelessly racist over time despite having changed society?AI will undoubtedly become an essential tool in the near future, but we must not forget the need to monitor the validity of its conclusions. Otherwise, we could always be slaves to the injustices of the past and bury the hopes of a future in which all people have the same opportunities.