The way we train AI is fundamentally flawed
A problem: A group of 40 Google researchers has identified a major cause for the common failure of machine-learning models. It’s called “underspecification” and it means we can’t tell if the machine-learning models we use today will work in the real world or not. That’s a real problem.
The evidence: The researchers looked at a range of different AI applications, from image recognition to natural language processing (NLP) to disease prediction. They found that underspecification was to blame for poor performance in all of them. The problem lies in the way that machine-learning models are trained and tested, and there’s no easy fix.
What it means: We need to be doing a lot more testing, but that won’t be easy.
Read the full story.
—Will Douglas Heaven
