5 ms·
"Google is still the biggest provider of high quality AI research papers in the world." I agree that they have at least one of the top AI research groups among
by 0xcafefood 4y ago
"Google is still the biggest provider of high quality AI research papers in the world."
I agree that they have at least one of the top AI research groups among the corporate players. But is that going to translate into business advantage?
IBM Research was (and is still to some extent) pretty solid, but did that actually translate into business advantage? The "Watson" that competed on Jeopardy was impressive for its time. But we can look back on how the attempt to commercialize that flopped pretty hard.
Note: I'm not saying Google has IBM levels of sclerosis, at least not yet.
But there's no ironclad law that a solid research division will always turn into something profitable.
- basementcat 4y agoIBM research has had multiple positive effects on its business (and industry). Just a few off the top of my head… * Giant Magnetoresistance (gigabyte hard drives) * Financial Swaps * Silicon on Insulator Technology * DRAM * Fortran * Copper metallization on silicon dies
- jeffbee 4y agoOutsiders just don't understand how ML contributes to Google's profits, because Google doesn't directly sell the ML to customers (like IBM's idiotic Watson product). But ML is plugged in everywhere: BERT (or its secret descendants) in Search, whatever unnamed inference system stands behind their Smart Reply and Smart Compose, the image segmenting and classification thing that makes searching Google Photos so easy. They even have a thing in Google Sheets that predicts what formula you are about to type. Their physical infrastructure from water pumps to cooling fans right down to CPU clock speed is controlled by machine-learned algorithms. They have a machine-learned compiler optimization pass that inserts explicit memory prefetches in key latency-critical function. ML is everywhere.