Andrew Ng

Led the Google Brain project that trained a neural network on 10 million YouTube thumbnails without labels, discovering features like 'cat detectors' purely from statistical patterns. Demonstrated that scale + unsupervised learning could discover meaningful representations — a key insight behind modern self-supervised learning.

Ng's most repeated analogy: like electricity 100 years ago, AI will transform every industry. Not a general-purpose tool but a foundational infrastructure that needs to be built into products and workflows everywhere. He has spent his career after Google Brain making this accessible — through Coursera, deeplearning.ai, and Landing AI.

Launched the data-centric AI movement in 2021. His argument: the industry is too focused on model architectures and too little on data quality, consistency, and labeling. For production AI systems, improving your dataset systematically often yields larger gains than trying a new model architecture.

Argued (2023–2024) that the biggest practical leap in AI is not better foundation models but agentic workflows — LLMs that plan, use tools, execute multi-step tasks, and iterate based on feedback. His 'four agentic design patterns' (reflection, tool use, planning, multi-agent) became a widely referenced framework.

Consistently pushes back on existential AI risk narratives (2023–present). His argument: overstating AI danger discourages beneficial applications, disproportionately harms developing nations that need AI most, and creates regulatory barriers that entrench large incumbents. He distinguishes between real, near-term harms (bias, misuse) and speculative far-future doom.