A study by the Model Evaluation and Threat Research found that programmers were 19% slower when using AI coding assistance tools than without. Despite offering help, AI caused programmers to allocate more time to reviewing outputs, prompting systems, and waiting for AI generations. Programmers accepted less than 44% of AI-generated tips without modification, losing time fixing errors produced by AI. Initially believing AI would speed up their tasks by 25%, they still thought it improved their efficiency by 20% after the study, highlighting a disconnect between expectations and reality.
A recent study showed that programmers were actually 19% slower when using AI coding assistance compared to coding without it, indicating inefficiencies in AI use.
Programmers using AI spent more time reviewing AI outputs and waiting for generations instead of actively coding, which hindered overall productivity significantly.
Collection
[
|
...
]