Scale Makes Things Interesting *and* Useful

My colleagues and I wear several hats: sometimes we're computer *scientists*, sometimes we're *mathematicians*, and sometimes we're software *engineers*. Let me invite controversy by reducing the distinction to "scientists and mathematicians strive for interestingness, engineers strive for usefulness." Academic computer scientists sometimes get stuck on problems that are moderately interesting, but not as useful as they'd like. Industry software engineers often end up building useful, boring things. I propose that the way out of this dichotomy is scale: doing really big things. Simple techniques, when applied to enormous quantities of data, yield useful results: e.g. Google spelling correction, or shotgun sequencing of genomes. Doing computation at large scales provides interesting systems challenges, e.g. MapReduce, or protein folding. I will suggest a number of examples where scale makes things interesting and useful, and discuss some details about where we've found this to be true at Google.