Peter Norvig is an American computer scientist known for co-authoring, with Stuart Russell, Artificial Intelligence: A Modern Approach, the standard textbook used in most university Artificial intelligence courses since 1995, and for a long research career spanning NASA and Google.
Career
Born in 1956 in New Jersey, Norvig earned a PhD in computer science from the University of California, Berkeley. Before joining industry, he was chief of the Computational Sciences Division at NASA Ames Research Center, where his group worked on AI systems for spacecraft autonomy and mission operations. He joined Google in 2001, eventually becoming Director of Research, a role in which he oversaw much of the company's early Machine learning research effort, including work that fed into products such as search ranking, translation and speech systems.
"The Unreasonable Effectiveness of Data"
In 2009, Norvig co-authored an influential essay with Alon Halevy and Fernando Pereira titled "The Unreasonable Effectiveness of Data," arguing that for many problems in language and vision, simple statistical models trained on very large datasets consistently outperformed more elaborate, theoretically motivated approaches trained on less data. The essay is frequently cited as an early articulation of the philosophy that later underpinned the Scaling laws observed in Deep learning and especially in Large language model development, where increasing Training data, parameters and compute has repeatedly produced better results than adding hand-crafted structure.
Education and later work
In 2011, Norvig co-taught an online introduction to artificial intelligence course at Stanford with Sebastian Thrun that drew over 100,000 enrolled students worldwide, a widely cited early moment in the massive open online course (MOOC) movement and a precursor to platforms such as Udacity and Coursera. He has continued to write and speak on AI education and on how the field's engineering practice has diverged from and converged with the theoretical picture presented in his and Russell's textbook, remaining, alongside Russell, one of the most widely read expositors of artificial intelligence as a coherent academic discipline even as its center of gravity shifted from search and logic toward statistical and neural methods over the three decades since AIMA's first edition. Successive editions of the book have been substantially rewritten to track that shift, expanding coverage of probabilistic reasoning, machine learning and, most recently, large-scale neural systems, while retaining the original rational-agent framing as the organizing thread across otherwise disparate subfields.