Semafor has made 310 contributions to the Pangram detector. The signatories take responsibility, the editors differ on what they authorize.
The survey published on 26 August by Semafor covers 310 fora published in the Wall Street Journal, the New York Times and the Washington Post, submitted to the Pangram detector. Verdict: 260 entirely human texts, about 40 partially generated, and ten 80% or more classified as machine produced.
The cases named illustrate embarrassment. Investor Stanley Druckenmiller, whose WSJ stand is 100 percent AI, claims the use without detour: he writes everything with AI for the same reason that he uses a calculator. Santiago Schnell, published on August 9 in the Washington Post and also ranked 100%, says he used ChatGPT to refine some arguments and correct grammar. A Jen Easterly forum on cybersecurity, published on August 2 in the New York Times, is detected as generated even as the daily bans AI from writing.
The three houses do not have the same doctrine. The New York Times strictly prohibits the use of AI. The Washington Post asks contributors to confirm that they have not used them. In the Wall Street Journal, the opinion page accepts it when the argument remains original — the most permissive position of the three, and probably the most difficult to verify.
Read the full article on the original source
Go to the source →