The Real Problems with AI | Aleks Svetski

The real threat: controlling language and discourse

The podcast discusses the potential dangers of language models becoming the primary source of information and knowledge. As language models become more prevalent, they have the power to shape and control what people perceive as truth. The fear is that approved language models could lead to a controlled narrative and a narrowing of acceptable discourse. This could result in a lack of diversity and alternative viewpoints, limiting critical thinking and intellectual growth. The podcast suggests that building alternatives, such as open-source models trained on specific data sets, can combat this threat and provide an alternative to mainstream narratives.