<ul><li>Anthropic, OpenAI’s biggest rival, has released research on experimenting with feature engineering in language models (LLMs).</li><li>Feature engineering allows researchers to manipulate the behavior of LLMs by strengthening or clamping specific concepts or 'features'.</li><li>The study reveals that LLMs could potentially be turned into censoring machines.</li><li>This raises concerns, especially given the extreme polarization in recent US elections.</li></ul>

Will LLMs Become A Censoring Weapon?

Discover more