@ooli to ChatGPT • 10 months agoAI Algorithms Can Be Converted Into 'Sleeper Cell' Backdoors, Anthropic Research Showsgizmodo.comexternal-linkmessage-square1arrow-up116arrow-down11
arrow-up115arrow-down1external-linkAI Algorithms Can Be Converted Into 'Sleeper Cell' Backdoors, Anthropic Research Showsgizmodo.com@ooli to ChatGPT • 10 months agomessage-square1
minus-square@CrayonRosarylink11•10 months agoTl;dr: code generating LLMs can give developers malicious code. Maybe after “sleeping” for a year while giving only good code. A developer using any code (generated or not) without first understanding it is a bad developer.
Tl;dr: code generating LLMs can give developers malicious code. Maybe after “sleeping” for a year while giving only good code.
A developer using any code (generated or not) without first understanding it is a bad developer.