fbpx Rationality and computational rationality in the age of AI. | University of Limerick arrow
Search icon

Rationality and computational rationality in the age of AI.

MACSI at the department of Mathematics and Statistics at the University of Limerick invites you to a seminar

Date: Thursday 26th September 2019, Room A2-002 @ 4p.m.

Speaker: Alessio Benavoli, Senior Lecturer at CSIS, http://alessiobenavoli.com/

Title: Rationality and computational rationality in the age of AI.

Abstract:   Artificial intelligence (AI) and Machine Learning research is revolutionising our lives and leading us to a world with self-driving cars, automated trading on stock exchanges etc.. Such applications require AI methods to be able to make rational choices and make robust decisions. Rationality means that an AI agent is assumed to take account of available information and uncertainty, potential costs and benefits, and to act consistently (logically) in choosing the best action. Robustness in decision making is required to both the known unknowns (the uncertainty in the world about which the agent can reason explicitly) and the unknown unknowns (unmodelled aspects).

However, the current methods are not sufficiently suited to address these issues. In this seminar,

I will briefly discuss the pitfalls of "general-recipe" machine learning (from a statistics point of view) and then introduce probabilistic (Bayesian) machine learning from a Von Neumann–Morgenstern perspective: "a rational AI agent maximises their expected utility".

I will briefly discuss about a general framework to model rationality, that encompasses different definitions of rationality from decision/market/game theory like no-Dutch-book, no-arbitrage, Nash-equilibrium theorems. Finally I will discuss computational rationality, that is about identifying decisions with highest expected utility, while taking into consideration the costs of computation in complex real-world problems in which most relevant calculations can only be approximated.

In other words, how can an AI agent make rational decisions when their computational power and computational time is limited? How can we formalize a theory of computational rationality? How does Nature solve this problem?

Further Information: If you have any questions regarding this seminar, please direct them to Dr Romina Gaburro ext 3193, email romina.gaburro@ul.ie or Clifford Nolan (061 202766), clifford.nolan@ul.ie).

Sfi logo