Absolutely Interdisciplinary 2023 Keynote: Blaise Agüera y Arcas

Blaise Agüera y Arcas is a VP and Fellow at Google Research who has been an active participant in cross-disciplinary dialogues about AI and ethics, fairness and bias, policy, and risk. With the recent release of the latest generation of large language models, it feels like the ground has shifted under our feet. What can interacting with these systems teach us about the nature of “intelligence”? In his keynote lecture, Agüera y Arcas will discuss AI and cognition as they relate to questions of norms, valuation, and sociality.

Social cognitive theory and AI

Social cognitive theory posits that learning takes place in a social context in which there is interaction and co-constitution between agents, their behaviours, and their environment. In this session, we will explore how multi-agent reinforcement learning (RL) can be used to examine this theory. RL models can formally test social cognitive dynamics by simulating agents and modelling behaviours that can help us better understand how social processes emerge, revealing implications for aligning AI with human values.

Speakers: William Cunningham, Joel Leibo, Nicolas Papernot (moderator)

Value alignment?

AI systems are increasingly being used for decisions that have significant consequences. Ensuring these systems align with human values can prevent unintended negative outcomes, ensure ethical decision-making, and help to build trust and accountability. Should AIs be forever aligned with human values? Will AIs always be treated as tools or sometimes as citizens? Do we trust our societies and civilizations—and AI—to evolve without centralized control?

Speakers: Blaise Agüera y Arcas, Gillian Hadfield (moderator), Richard Sutton

Large language models

Recent advances in large language models (LLMs) are transforming the way we communicate with each other and interact with information. Educators, in particular, from the primary to the postsecondary level, are now presented with critical new opportunities and challenges. How will LLMs be used as a tool, whether for teaching or “cheating”? How has this dynamic played out in the history of technology (e.g. calculators, computers)? This session will feature a mix of researchers and practitioners debating the merits and hazards of LLMs in the classroom.

Speakers: Ashton Anderson (moderator), Lauren Bialystok, Paolo Granata

The reward hypothesis

Almost 20 years ago, AI research pioneer Richard Sutton posited the reward hypothesis: “That all of what we mean by goals and purposes can be well thought of as maximization of the expected value of the cumulative sum of a received scalar signal (reward).” Since then, advances in reinforcement learning have demonstrated that complex behaviours can emerge from artificial agents guided by scalar reward. Humanists and social scientists are starting to see the utility of the hypothesis as a claim about humans, although many disagree. The question remains, is the reward hypothesis of reinforcement learning a good model for understanding human behaviour and values? How far can it go? Can it guide normative decision-making for individuals and groups? For societies?

Almost 20 years ago, AI research pioneer Richard Sutton posited the reward hypothesis: “That all of what we mean by goals and purposes can be well thought of as maximization of the expected value of the cumulative sum of a received scalar signal (reward).” Since then, advances in reinforcement learning have demonstrated that complex behaviours can emerge from artificial agents guided by scalar reward. Humanists and social scientists are starting to see the utility of the hypothesis as a claim about humans, although many disagree. The question remains, is the reward hypothesis of reinforcement learning a good model for understanding human behaviour and values? How far can it go? Can it guide normative decision-making for individuals and groups? For societies?

Speakers: Julia Haas, Gillian Hadfield (moderator), Richard SuttonJulia Haas, Gillian Hadfield (moderator), Richard Sutton

Machine learning in the workplace

How will recent advances in AI change the nature of work? This session will feature a discussion between economist Daniel Rock and computer scientist Frank Rudzicz. Rudzicz will discuss his experience implementing ML tools into hospital workflows, while Rock highlights his work on how generative AI tools are likely to impact jobs.

Speakers: Avi Goldfarb (moderator), Daniel Rock, Frank Rudzicz

AI and creativity

How can we re-conceptualize creativity, whether human or non-human, in light of the latest advances in AI and our interactions with it? How do new technologies impact our conceptions of self, language, expression, and art? Incorporating tools and insights from science and technology studies, literary criticism, and creative practice, this session will turn a humanities lens on the crucial sociotechnical problems of the current moment.

Speakers: Polly Denny, N. Katherine Hayles, Avery Slater (moderator)