Welcome to Teeny-Tiny Castle

AIRES at PUCRS is making teeny-tiny_castle available to the public — a code repository aimed at giving researchers interested in working on AI ethics and safety issues the tools they need, built in Python.

In this repository, you’ll find notebooks and scripts showing how to build and use tools to address certain questions we work on in AI ethics and safety (e.g., interpretability, sustainability, fairness, robustness). You’ll also find an introductory Python+ML ( - Machine Learning - ) course to help you get started if you’re not familiar with the language. The Intro to ML section was designed to give newcomers a grasp of some of the tools and abstractions that underpin ML.

All the requirements and step-by-step instructions can be found in teeny-tiny_castle. There you’ll find a series of tutorials, examples, and tools for working with topics related to:

AI Ethics (Guidelines, Governance, Regulation, R&D).

Sustainability in large-model development (how to measure and quantify the carbon footprint of ML-trained models).

Interpretability and robustness in computer vision (tools for XAI and Adversarial ML).

Interpretability and Robustness in NLP (LIME for NLP, prompt-engineering playgrounds, text-mining examples).

Interpretability in classification and prediction with tabular data (how to explain divergent classifications from ML-developed models).

Fairness in machine learning (how to measure, quantify, and remedy algorithmic discrimination).

Systemic security (how organizations should coordinate to protect and preserve their information infrastructure).

This repository is a work in progress, being built as part of the RAIES initiative (Network for Safe and Trustworthy Artificial Intelligence), a project supported by FAPERGS (Foundation for Research Support of the State of Rio Grande do Sul). If you’re interested in collaborating on the project, contact us!