Giuseppe Attanasio

Postdoctoral Researcher

Ciao!

I am a Postdoctoral Researcher in the SARDINE group at Instituto de Telecomunicações, Lisbon, headed by scientific director André Martins. My research interests lie at the intersection of interpretability, fairness, and safety. I previously joined the MilaNLP group where I was working with Debora Nozza and Dirk Hovy.

I am keeping my latest updates in News.

I care about the impact of societal biases infused in AI technologies, about which I gave talks and published modest science. I believe education and associationism play essential roles. I co-founded a student team and an online community to connect passionate people. I also believe deep understanding is rooted in hands-on exercise: our tutorial on implementing a Transformer from scratch in JAX is an example.

Interests

Interpretability
Fairness
Safety

Education

Doctor of Philosophy (Ph.D.), 2022

Politecnico di Torino
MEng in Computer Engineering, 2018

Politecnico di Torino
BSc in Computer Engineering, 2016

Politecnico di Torino

Talks

ferret: A Framework for Benchmarking Explainers on Transformers

This talk provides a comprehensive overview of ferret, a novel python library for benchmarking interpretability techniques on …

Mar 23, 2023 12:00 AM Turin, Italy

Giuseppe Attanasio, Eliana Pastor, Chiara Di Bonaventura, Debora Nozza

ferret: A Framework for Benchmarking Explainers on Transformers

From Recurrent Models to the advent of Attention

All recent language technologies, such as BERT, GPT3, or ChatGPT, rely on complex neural networks called Transformers. But, what …

Feb 15, 2023 6:30 PM eDreams ODIGEO Tech Hub

From Recurrent Models to the advent of Attention

Social Biases in LMs: Why they are there, how do we measure and mitigate them

In the blooming age of language technologies, the impact of social biases encoded in our models has surged to unprecedented levels. In this half-talk, half-discussion, we go through fundamental definitions of social bias in computer systems, specifically language models and social harm.

Oct 28, 2022 1:00 PM — 1:00 PM

Social Biases in LMs: Why they are there, how do we measure and mitigate them

See all events

Projects

Voice Bot

A Telegram bot to transcribe any voice note with speech AI models.

Risorse per la Lingua Italiana

An online community to discuss about and build computational linguistic tools in Italian.

Yet Another (JAX) Transformer

An annotated tutorial to build and train a Transformer Language Model.

MALTO Student Team

MAchine Learning At poliTO (MALTO) is a university-funded student team. The field of data science is becoming increasingly more important in our everyday life. In particular, at Politecnico di Torino, we strive to become national and international leaders in the field.

Hypa Watering System

Hypa is an automated watering system I have built to keep my green friends hydrated. My PhD working activities are mainly based in Turin. However, I am from Castrovillari, a twenty-thousand people town in southern Italy, where I love to spend my holidays, close to my family.