Andrew Ng

Founder — DeepLearning.AIAdjunct Professor — StanfordManaging General Partner — AI Fund

Profile

If you’ve taken an online machine learning course, there’s a decent chance Andrew Ng taught it. His 2011 Stanford machine learning class on Coursera enrolled over 100,000 students in its first run — a number that seemed absurd at the time and helped invent the MOOC category. He co-founded Coursera with Daphne Koller shortly after. More people have learned ML fundamentals from Ng than from any other single human, full stop. If you’re a developer who picked up neural networks from a video lecture in the last decade, he’s probably in your lineage.

Before the teaching empire, Ng co-founded Google Brain in 2011 alongside Jeff Dean and others — the team that trained a neural network on YouTube thumbnails and famously found neurons that responded to cats. He then ran AI at Baidu as Chief Scientist from 2014 to 2017, building out one of the earliest large-scale industrial deep learning groups outside the US. He’s been an adjunct professor at Stanford the whole time, with CS229 and CS230 lectures freely available and still assigned reading for anyone getting serious.

Today he runs a small empire focused on getting AI into more hands. DeepLearning.AI produces the Deep Learning Specialization, the Generative AI specializations, and short courses with partners like OpenAI, Anthropic, and LangChain. AI Fund is his venture studio that spins up AI companies from scratch. Landing AI focuses on computer vision for manufacturing. And The Batch is his weekly newsletter — one of the better signal-to-noise reads in the field.

Ng’s stance is consistently pragmatic and builder-friendly. He’s skeptical of AI doom narratives, vocally pro-open-source, and relentlessly focused on “what can you actually build with this.” He famously called AI “the new electricity” — a phrase that’s been mocked and quoted to death but captures his worldview. For someone learning AI to build things, he’s the most useful person to follow: no hype, no catastrophism, just a steady drumbeat of “here’s what works, here’s how to ship it.”

Books

Machine Learning Yearning A free book on how to structure machine learning projects — practical advice on error analysis, train/dev/test splits, and debugging. Focuses on the engineering judgment calls, not the math.

Key Articles & Papers

Building High-level Features Using Large-Scale Unsupervised Learning 2012 — The Google Brain 'cat neuron' paper. Unsupervised training on YouTube thumbnails produced a neuron that fired for cat faces — an early, vivid demonstration of representation learning at scale. Deep learning with COTS HPC systems 2013 — Showed you could train billion-parameter networks on commodity GPU clusters rather than giant CPU farms. Quietly important — helped make large-scale deep learning affordable outside of Google. Deep Speech: Scaling up end-to-end speech recognition 2014 — Ng's Baidu team showed an end-to-end neural approach beating classical pipelines for speech recognition. A template for the 'stop hand-engineering, throw data at it' pattern that came to define the decade. What Artificial Intelligence Can and Can't Do Right Now 2016 — Ng's Harvard Business Review piece giving non-technical leaders a mental model: if a task takes a human under a second of thought, AI can probably automate it. Simple, useful framing. AI Transformation Playbook 2018 — A five-step playbook for how companies should adopt AI. Written for executives but useful for engineers who want to understand why AI projects fail organizationally. Data-Centric AI 2021 — Ng's pitch that improving data quality often beats tuning model architecture. Kicked off a minor movement and a lot of tooling — an antidote to pure model-worship.

Controversies

Ng has been openly skeptical of AI existential risk narratives, comparing worry about superintelligent AI to “worrying about overpopulation on Mars.” This has put him repeatedly at odds with figures like Eliezer Yudkowsky, Geoffrey Hinton, and Yoshua Bengio. He’s also been a vocal critic of heavy-handed AI regulation — particularly California’s SB 1047 — arguing that compute thresholds and liability regimes will kill open source and hand the field to a few large labs. Supporters call this refreshingly grounded; critics accuse him of dismissing serious safety concerns from people who built the technology he teaches.