Why AI? A Deep Dive into Neural Networks, Transformers, and the Governance of Artificial Intelligence

The Naked Scientists unlocks the story of artificial intelligence, tracing its roots from brain-inspired networks to modern large-scale models like ChatGPT. The episode blends expert history with present day debates on safety, governance, and the future of work, featuring Geoffrey Hinton, Mike Wooldridge, and Dame Wendy Hall among others.

• Key insights into backpropagation and how neural networks learn
• Evolution from symbolic AI to deep learning and the transformer era
• Optimism about AI balanced with real concerns about misuse and regulation
• Calls for international AI safety standards and independent auditing

Introduction and partnership

The podcast introduces a new series in collaboration with Frontiers, framing the central question, why AI, and outlining how the conversation will connect science, policy, and technology at Davos.

Foundations of AI: from neurons to backpropagation

Geoffrey Hinton discusses the inspiration from brain networks, the early challenges of learning rules, and how backpropagation emerged as a practical algorithm. The host and Hinton walk through a conceptual tour of input and output layers, hidden layers, and how incremental changes to connection strengths converge toward correct recognition tasks such as digits or images, highlighting the efficiency gain from backpropagation versus one-by-one trial and error.

Two AI traditions and the turning point to deep learning

The discussion covers the 1980s to 1990s split between neural networks and symbolic AI, the rise of data and compute power, and the pivotal role of GPUs for accelerating mathematical heavy lifting. OpenAI and the transformer architecture are presented as the leap that enabled training unprecedentedly large models on vast data.

Scale, transformers, and widespread adoption

Transformers, large-scale data, and the web together catalyzed the birth of modern AI systems, culminating in products like ChatGPT. The narrative connects core science with concrete everyday uses, from image and speech recognition to personal assistants.

Optimism, concerns, and governance

The episode shifts to risk assessment: while the technologists are optimistic about capability, insiders warn of misuse, regulatory gaps, and the potential disruption to labor and mentorship. The conversation includes the theme that governance and oversight are essential to realizing AI’s benefits responsibly.

Regulation, oversight, and the path forward

Christina Criddle and Dame Wendy Hall introduce the tension between a rushed frontier and the need for guardrails. Hall advocates cooling hysteria while proposing independent multinational oversight, possibly via a UN-backed authority, to audit safety processes and hold companies liable for harm. The discussion also touches on global dynamics between the US and China in setting standards for AI safety.

Closing perspectives

The podcast closes with reflections on journalism in an AI-rich era, the importance of trusted, human-driven reporting, and a look ahead to further explorations on topics like geoengineering, bringing the Davos-connected Frontiers Science House into focus.