AI writing has already begun to appear on the opinion pages

The majority of guest columns in the opinion pages of major newspapers like The Wall Street Journal, The Washington Post, and The New York Times are still written by humans, according to a recent analysis by Semafor. However, large language models have started to appear in these publications, albeit on a limited scale. Out of 310 guest submissions to the three publications over the past month, Pangram, an AI detector, identified 10 as at least 80% AI-generated.

Another 40 articles were partially AI-generated. Pangram's accuracy in detecting AI-written text is generally high, though some users have reported false positives. The controversy surrounding AI in journalism gained momentum after billionaire Stanley Druckenmiller used AI to write an op-ed criticizing Treasury Secretary Scott Bessent's interventions in the bond markets.

The op-ed received a 100% AI score from Pangram. Druckenmiller defended his use of AI, stating that he now relies on it for writing like he uses a calculator for math problems. The New York Times has a policy prohibiting the use of AI in developing and drafting guest essays, while The Washington Post requires guest writers to confirm their submissions were not AI-generated.

Stanford research suggests that detection tools can sometimes flag work by non-native English speakers, leading to concerns about bias. Gigot of The Wall Street Journal noted that AI is "a fact of modern life," but emphasized that the important factor is whether the published content reflects the author's original argument and if they have the credibility to make it.

The analysis suggests that the low usage of AI-generated content in these publications may be due to concerns about quality and potential embarrassment. Despite the potential for AI models to improve over time, the current state of large language models is still considered a "lowest common denominator of human discourse."