Nate Meyvis

When I worked in machine learning every day, I found a close analogue of this to be very true. A slightly better architecture tended to matter much less than getting better data. Or, at least, once I hit on a reasonable model architecture, I tended to do much better by cleaning and gathering more data than by optimizing the architecture.

GROSS: We’re certainly not saying that horsepower and IQ don’t matter. They certainly do, but there is a clearing bar at which point, for most roles, I think people tend to overvalue it and don’t realize the logarithmic nature of the curve and assume it’s linear. But I actually think, earlier than many people realize, you want to switch to caring a... See more