Why AI is incredibly smart -- and shockingly stupid | Yejin Choi
Mar 22, 2024
auto_awesome
Computer scientist Yejin Choi discusses the limitations and funny fails of AI in common-sense reasoning. She explores the benefits of training AI on human norms and values. The podcast delves into the power, pitfalls, and ethical dilemmas of AI, suggesting ways to democratize and humanize it. Innovative solutions like Paycom's Betty and Zbiotics are also highlighted.
Large AI models raise concerns about power concentration and environmental impact, requiring scrutiny and ethical considerations.
Teaching AI common sense values is crucial for aligning AI behavior with human understanding and values.
Deep dives
The Challenges of Extreme Scale AI Models
Large AI models trained on extreme scales raise concerns about concentration of power in tech companies, limited scrutiny due to cost, and environmental impact. Questions arise about AI safety without common sense and the effectiveness of brute-force scaling for AI training.
The Importance of Common Sense in AI Development
Common sense is a critical challenge in AI, impacting decision-making and ethical considerations. The analogy of dark matter highlights the unseen norms influencing language usage. Teaching AI common sense values and norms is essential for aligning AI behavior with human understanding and values.
Innovative Approaches for Teaching AI Common Sense
Innovations in data and algorithms are crucial for enhancing AI training. Emphasizing human-developed data and feedback for AI training can address biases in raw web data. New algorithms like symbolic knowledge distillation aim to distill common sense knowledge and promote transparent and human-inspectable AI models for ethical and sustainable AI development.
Computer scientist Yejin Choi is here to demystify the current state of massive artificial intelligence systems like ChatGPT, highlighting three key problems with cutting-edge large language models (including some funny instances of them failing at basic commonsense reasoning.) She welcomes us into a new era in which AI is becoming almost like a new intellectual species -- and identifies the benefits of building smaller AI systems trained on human norms and values. (Followed by a Q&A with head of TED Chris Anderson)