Abstract: I will present some joint work with Amnesty International, leveraging crowd-sourcing on two human-rights cases :
(i) Online abuse against women on Twitter. This is the first hand-in-hand collaboration between human rights activists and machine learners. On a technical front, we carefully curate an unbiased yet low-variance dataset of labeled tweets, analyze it to account for the variability of abuse perception, and establish baselines, preparing it for release to community research efforts. On a social impact front, this study provides the technical backbone for a media campaign aimed at raising public and deciders’ awareness and elevating the standards expected from social media companies.
For more details see https://decoders.amnesty.org/projects/troll-patrol/findings
(i) Our recent work on Super-Resolution and its potential for NGOs relying on cheap low-resolution imagery to monitor human rights and the environment from space. Multi-Frame Super-Resolution (MFSR), a technique to merge several low resolution images to one image of higher quality, seems to be a promising approach to artificially enhance images, in particular remote sensing data. Our work suggests that neural networks can learn data-driven representations of multiple frames to de-alias low-resolution signals and effectively enhance large amounts of earth observation data, at scale.
See this blog to learn about our vision for Super-Resolution technology:
Bio: Alfredo joined Element AI as a Research Engineer in the AI for Good lab in London, working on applications that enable NGOs and non-profits.
He is one of the primary co-authors of the first technical report made in partnership with Amnesty International, on the large-scale study of online abuse against women on Twitter from crowd-sourced data. He's been a Machine Learning mentor at NASA's Frontier Development Program, helping teams apply AI for scientific space problems. More recently, he led the joint-research with Mila Montreal on Multi-Frame Super-Resolution, that was awarded by the European Space Agency for their top performance on the PROBA-V Super-Resolution challenge.
His research interests lie in computer vision for satellite imagery, probabilistic modeling, and AI for Social Good.
Prior to joining Element AI, he was a Senior Data Scientist in Digital Shadows, specializing in cyber-security and digital risk management, and a consulting Data Scientist in Microsoft's Xbox EMEA team, where he also collaborated with Microsoft Research Cambridge.
He has been a research scientist with the Department of Statistical Science at University College London, working on probability models to better understand ordinal data coming from surveys, and later on the interface of Machine Learning and Signal Processing to detect faults in the low-voltage power-line grid. During his time with UCL his team won the first data challenge competition organised by the Royal Statistical Society, for which he designed and developed his team's algorithm for the analysis of resting f-MRI time-series data.
He earned his MSc in Artificial Intelligence from the University of Edinburgh, and his PhD in Machine Learning from the University of Sheffield under the supervision of Professor Neil Lawrence. His PhD research led to contributions in probability methods for dimensionality reduction of data and developed methods for gene-expression time-series to discover genetic factors of disease.