Data Scientist at Boston Consulting Group (BCG) • Tech Author • ML Engineer • Pharmacist
A common issue with most educational content on the responsible use of AI is that they are too theoretical and text-heavy. While these resources do contain valuable ideas, the format can be less engaging and digestable for practitioners. In the Responsible AI masterclass I conducted recently for learners from the IBFSG - The Institute of Banking & Finance Singapore - Technology in Finance Immersion Programme (TFIP) program, I shifted the focus towards practical, hands-on learning within the context of financial services. I introduced three open-source libraries, each targeting a different aspect: fairness and transparency diagnosis, debiasing algorithms, and guardrails for large language models (LLMs). ① 𝐕𝐞𝐫𝐢𝐭𝐚𝐬 (by Monetary Authority of Singapore (MAS) and finance industry partners): Provides us with the capability to easily assess fairness and transparency of machine learning models ② 𝐀𝐈 𝐅𝐚𝐢𝐫𝐧𝐞𝐬𝐬 360 (by IBM): Allows us to readily detect and mitigate bias in machine learning models across the AI application lifecycle ③ 𝐍𝐞𝐌𝐨 𝐆𝐮𝐚𝐫𝐝𝐫𝐚𝐢𝐥𝐬 (by NVIDIA): Enables us to easily add programmable guardrails to LLM-based conversational systems I have publicly released the lesson Colab notebook on the Veritas toolkit, available here: https://lnkd.in/geRhwwKP My thoughts on 𝐕𝐞𝐫𝐢𝐭𝐚𝐬: - Great starting point for understanding and implementing the concepts of fairness evaluation (easier to get started than the AIF360 toolkit which has poor integration with sklearn) - However, it is >1 year since the last commit (pretty much stale) and the documentation can be much improved (prompting me to develop the Colab notebook) I hope this is useful for anyone conducting future lessons on responsible AI. I’m also eager to hear from others in the field: Are there other resources you've found effective for applying responsible AI principles in practical settings? Link to GitHub repo: https://lnkd.in/gxmgYdT8 #datascience #machinelearning #responsibleAI #generativeai #bcg #artificialintelligence #rai RISE by BCG U
Kenneth Leung to teach with learners in mind is indeed important. Kudos 👏
Thanks for sharing! This is very useful! Bookmarking this for my future learners!
Data Scientist at Boston Consulting Group (BCG) • Tech Author • ML Engineer • Pharmacist
1moAnd here's the Colab notebook for NeMo Guardrails: https://colab.research.google.com/drive/1a5ysrAJPSfAszPAkJPTzqP0ZYlDFo8Zb