Rich Sutton's March 2019 essay arguing that the biggest lesson from 70 years of AI research is that general methods leveraging computation ultimately prove most effective, by a large margin. Sutton walks through prominent cases — computer chess, Go, speech recognition, and computer vision — where approaches built on human knowledge were outdone by search and learning methods that scaled with computing power. He attributes the pattern to the steadily falling cost of computation and warns that relying on human knowledge tends to complicate methods in ways that limit scaling.
#AI research#essay#compute#search and learning#Rich Sutton#machine learning
Stanford HAI's annual flagship report tracking AI progress across research, the economy, education, policy, and public opinion. The 2025 edition compiles data on model capability, training costs, industry investment, and workforce effects, with downloadable charts and datasets.
Why I recommend it: Free to download from Stanford HAI. It is a self-published report from a university institute; figures are sourced within the report, but some industry-investment and capability numbers rely on data supplied by the companies being measured. Treat headline rankings as the report's own synthesis, not neutral fact.
The Tony Blair Institute's hub explaining compute — the hardware, software and infrastructure stack that stores, processes and moves data at scale — and arguing that access to it now determines which countries and public services can use AI at all. Collects its work on infrastructure, digital skills, regulation and international collaboration.
Why I recommend it: Useful for the plain definition and for seeing how governments are being asked to think about this. It is advocacy, not neutral analysis: the Institute campaigns for rapid state adoption of technology and is funded in part by technology donors, so it rarely dwells on the land, water and electricity costs that the same build-out imposes locally. Read it alongside the data-center cases on our impacts page.
Trillions in compute commitments come due in 2027–2028. The Reset Wall reveals how the AI boom breaks, and when.
From the site: Trillions in compute commitments come due in 2027–2028. The Reset Wall reveals how the AI boom breaks, and when.
Why I recommend it: A clear-eyed look at where the AI build-out may hit a financing and infrastructure wall — useful context for anyone advising job seekers or founders betting on the sector.
