María Alonso (Pontevedra, 29 years old) is reluctant to use ChatGPT in her statistical research work, except if she has to adapt a text. In fact, she is in charge of creating statistical models, better than those behind artificial intelligence (AI), to interpret certain specific data: “If an artificial intelligence tries to predict, for example, the direction of the wind or the waves, it will do it wrong. It does not take into account whether the data is on a circle and not on the real straight line.” Alonso is not looking for a model that makes a prediction, but rather to mathematically demonstrate that the predictions are well made. Thanks to her work, she has won the Vicent Caselles Award from the BBVA Foundation and the Royal Spanish Mathematical Society (RSME) which encourages the talent of young mathematicians.
During Spanish national football matches or basketball broadcasts, Alonso puts statistics to work in her head: she tries to calculate the probability of a team winning or not. She downloads the data that interests her and tries to predict what will happen in the future. She takes into account, for example, the number of minutes a player plays or the number of shots on goal. Alonso studied mathematics at the University of Santiago de Compostela, where she is now a professor, and also completed a doctorate in Statistics and Operational Research.
Ask. How do you use statistics in your research?
Answer. We are working on predicting wave direction and height on the Costa da Morte in Galicia, a very dangerous spot. It is important to model maritime behaviour based on many variables: salinity, water temperature, wind direction, wind speed. If you have these weather conditions, you can say when boats should not set sail.
P. Are normal techniques not suitable for that type of data?
R. Exactly. They are designed for variables defined on the real line, but the angles are not defined on the real line, but on the circumference, and the support itself is different. The project consists of designing new techniques to analyze variables for this type of data. Being able to see how they relate to each other or predict a variable based on the information given by others.
P. Does artificial intelligence take into account such specific data?
R. Classical statistics or models of machine learning The classic or artificial reference models that exist up to now do not do this. We start from scratch: we build statistical models that analyse them correctly.
P. Can you give some examples?
R. I have found scientific articles in the field of biology where they wanted to study escape strategies in animals, in which direction an animal escapes when it is pursued by a predator. The authors used usual statistical techniques, and drew conclusions that are erroneous because it is built around the mean of a random variable that is defined on the real line. What I do is build similar models that take into account that you are not on the real line, that you are on a circle, and then study that these models work correctly.
P. The models he works with also help in fields such as psychology and neurology.
R. In neurology, we have research with macaques in which they wanted to study the peaks of neuronal activity in the primary cortex of the brain depending on the direction of a stimulus. They had the macaques sitting and sedated and they produced a visual stimulus from different directions, to study whether there was more neuronal activity in specific directions. We applied certain techniques in which we saw that, depending on the direction of the stimulus, the neuronal activity was greater or lesser, and the variance of that neuronal activity also changed with the direction of the stimulus.
P. What is artificial intelligence doing for mathematics now that couldn’t be done before?
R. What is behind artificial intelligence is mathematics. I find it difficult to visualize how it could take the job of mathematicians. Our job is really to think. Instead of spending ten hours thinking about how to program something, I can call ChatGPT. In programming it is a paradigm shift, but it is not the same in basic mathematics. When you do mathematics you are trying to see things that no one has seen. If you try to do original things, it is very difficult for artificial intelligence to replace you, because it only repeats things that it has already seen. Mathematics grows, every day something new is discovered.
P. Is it harder for AI to replace creative jobs?
R. Everything that has to do with creativity is what artificial intelligence will not be able to do. It will never do something completely different from what it has been trained to do.
P. Does statistics help reduce model biases?
R. If I only train the model with quotes from famous male mathematicians, then it will only respond with things related to men, because I have given it that data. It is not that statistics or artificial intelligence is biased, it is society itself that is biased and we inherit our own biases. The sad part of applying statistics is that many people try to manipulate the results through this kind of things. You have to know that there is a bias in the data and try to correct it.
P. Have you seen any bias against women?
R. If you ask artificial intelligence to draw a person who is dedicated to researching mathematics, all the drawings are men. A very clear stereotype of an older man with glasses, a mathematical profile or profiles more related to science and technology. In that case, artificial intelligence is not going to give you back a young female mathematician.
P. What would you have studied if you hadn’t studied mathematics?
R. Teaching, I always liked children. I enrolled in mathematics thinking of being a teacher. I would have also opted for engineering, but perhaps because of the social perception that people told me “for engineering you have to be very clever” that women are studious, but we are not as clever, so I never even considered engineering for fear of not being worth it. My classmates at school told me that I was not intelligent, that I studied a lot and that was why I got good grades. In any case, I am glad because I really like the career I chose.
P. Are there many young women in the sector?
R. In statistics, the percentage of women is higher than in pure mathematics. In conferences and in research careers, the percentage is quite low. In statistics, it is higher because it is associated with a more applied profile and not with that of a mathematical genius. Women are afraid of not being capable.
P. What do you think is needed to motivate women?
R. They have to start from an early age to remove the biases that exist, those that say that girls are very hard-working and boys are very intelligent. Those comments affected me a lot when I was little. Eradicate them and try to show girls female role models in mathematics and science in general. It is not about forcing girls to study mathematics, but if you like it, let them know that you can and that you are very capable.
P. Can artificial intelligence be a hook?
R. Yes, when I started my degree, nobody wanted to study mathematics, they thought that it was only good for being a teacher. My family and friends told me that I was throwing away my entrance exam grade. Now it’s the opposite, companies want more mathematics and the cut-off marks are very high. This is largely due to the boom of the application of AI or machine learning and its use in companies, which are no longer looking for an educational profile, but rather a more technological one.
P. Do you think there should be a subject on artificial intelligence in schools?
R. I think it’s very hasty because it’s difficult to understand artificial intelligence. People sometimes want to go very fast and become experts in it overnight. machine learning without having to study mathematics. That is impossible, without the foundation you will not be able to understand what is behind it: it is like wanting to run without knowing how to walk. To understand how artificial intelligence works you have to be able to think logically, what mathematics does.
P. Will AI ever match human responses?
R. It depends. Everything that has to do with originality is going to be difficult to match. We are already seeing this in more routine things. It is trained with newspaper headlines, or with people’s opinions, which do not have to be true. Humanities professions, such as philosophy, are very necessary to analyze what is happening. I am always afraid that people will think that artificial intelligence is intelligent, that what it says is correct.
P. What can a philosopher do?
R. It can advise you on how much attention we should pay to artificial intelligence. Where is the limit at which an artificial intelligence can determine different things. If in medicine artificial intelligence tells you that you have to operate on one person before another, are you going to listen to it immediately or are you going to stop and consider both cases? AI can replace mechanical aspects, but it will always be necessary to have a human vision behind it, so that we are able to see what is right and what is wrong without being confused by the machine.
You can follow THE COUNTRY Technology in Facebook and X or sign up here to receive our weekly newsletter.