Amazon has revised its guidelines to steer its engineers toward purposeful AI use based on expertise, judgment, and the creation of new solutions. Terk Terkowitz, vice president and distinguished engineer at Amazon, says that deriving no benefit from AI is a finding to document and explain, not a failure. He argues that counting tokens and lines of code is a poor productivity metric that can increase costs and complexity. Amazon previously ended the internal Kirorank leaderboard, which encouraged employees to run agents on meaningless tasks, while retaining some AI adoption metrics and recommending comparisons between the least and most active users.
Source: Read the original article


