#74 Dr. ANDREW LAMPINEN - Symbolic behaviour in AI [UNPLUGGED]

Fairness

The discussion on fairness in AI highlights the complexity of defining and achieving ethical standards in current systems. Andrew Lampinen emphasizes the importance of systems that can reach conventional agreements about meanings, akin to human legal systems that rely on judges to interpret rules in context 1. He warns against simplistic solutions like imposing static rules, as fairness is an evolving concept that requires nuanced understanding and adaptability.

What you need is systems that can sort of come to conventional agreements about meanings.

--- Andrew Lampinen

Keith Duggar adds that AI should incorporate human judgment to navigate the vagueness inherent in ethical decision-making 1.

   

Ethical Training

Ethical training methods for AI focus on preventing the replication of human biases and negative behaviors. Andrew Lampinen discusses the challenge of training AI to understand human behavior without adopting its worst traits, emphasizing the need for carefully curated datasets 2. He advocates for systems with epistemic humility, allowing them to evolve ethically as human understanding progresses 3.

It's important for our artificial intelligence systems to be able to develop a framework of ethics.

--- Andrew Lampinen

Tim Scarfe likens the bad parts of human behavior to a storm, questioning whether such elements are necessary for a balanced AI 3.