XAI has witnessed unprecedented growth in both academia and industry in recent years (alongside AI itself), given its crucial role in supporting human-AI partnerships whereby (potentially opaque) data-driven AI methods can be intelligibly and safely deployed by humans in a variety of settings, such as finance, healthcare and law. XAI is positioned at the intersection of AI, human-computer interaction, the social sciences (and in particular psychology) and applications.
Overall, XAI is increasingly part of all AI policies on ethics, trustworthiness and safety of AI. This seminar series focuses on all aspects of XAI, ranging from methods to applications.
At the irruption of the COVID-19 pandemic, a group of researchers put together the SPIRA Project to collect, analyze and automatically detect respiratory insufficiency from voice audios of Brazilian Portuguese using cellphones. It was, by construction, a multidisciplinary initiative, consisting of computer scientists, medical doctors, speech therapists and linguists, which was conceived to study both big data analysis of audio via several machine learning methods, as well as small data analysis of acoustic properties of patients and control voice signals. Instead of detecting COVID-19 infection, the project aimed at detecting the main condition associated with the disease that forces people to be hospitalized, taking the point of view of patient triage. This view was motivated by the fact that SARS-CoV-2 infection was commonly associated with silent hypoxia, a condition in which there is a low oxygen saturation in the blood but the patient does not feel breathless. The project obtained 96% of accuracy in respiratory insufficiency detection using a Transformer neural network, plus a detailed audio analysis of fundamental frequency related parameters in Brazilian with COVID-19 and a characterization of pauses as a speech biomarker for COVID. Audio-processing techniques developed by the project obtained first prizes in competitions for COVID-19 detection in cough audios and in speech emotion recognition.
Marcelo Finger has a BSc in Electronic Engineering from Universidade de São Paulo (EP-USP 1988), MSc in Foundations of Advanced Information Technology from the Imperial College of Science, Technology and Medicine (1990) and PhD in Computing from Imperial College of Science and Technology, University of London (1994). He ha held visiting positions in departments of Computer Science at Universitée Paul Sabatier - Toulouse (2011) and Cornell University (2012-2013). He is currentely Professor of Computer Sciece at the Department of Computer Science, Institute of Mathematics and Statistics at the University of São Paulo, Principal Investigator at the USP-Fapesp-IBM Center for Artificial Intelligence (C4AI), where he coordinates the NLP2 group on natural language processing in Portuguese. He is part of the editorial board of the following journals: South American Journal of Logic, São Paulo Journal of Mathematical Sciences and has been a guest editor at Theoretical Computer Science and Anals of Mathematics in Artificial Intelligence. He as been acting as an expert in the Computer Sciece, focusing on Logics and Dedecutive-Probabilistic Reasoning, acting as a researcher in the following subjects: logic, artificial intelligence, Digital Humanities and computational linguistics.