AI Evaluation: From Traditional Paradigms to New Avenues in the Era of General-Purpose AI
Dr Lorenzo Pacchiardi is a Research Associate at the Leverhulme Centre for the Future of Intelligence here at Cambridge, where he develops frameworks to evaluate the cognitive capabilities of LLMs, as well as their ability to perform data science tasks.
His research background spans wide. He completed his PhD in *Statistics and Machine Learning at Oxford, focusing on Bayesian simulation-based inference, generative models, and probabilistic forecasting. He has since worked on *detecting deception in LLMs through the MATS programme.
In parallel, Lorenzo has been active in AI policy and advisory work: he contributed to the EU AI Act by developing technical standards at the Future of Life Institute, co-founded the Italian AI policy think tank CePTE, and is an advisor to RAND on AI evaluation.
In this talk, he’ll be sharing some of his recent work on the landscape of AI evaluation. We’ll then open the floor for discussion and Q&A.
Find out more about his work here: https://www.lorenzopacchiardi.me/