As more people use GPT-4, there are more interesting use-cases of the new technology being uncovered.
A Twitter user has discovered that GPT-4 can know who the author of a particular piece is based on just a few paragraphs of text. “GPT-4 is able to infer authorship from a passage of text based on style and content alone,” wrote Twitter user @vagabondjack. “Given the first four paragraphs of the March 13, 2023 @stratechery post on SVB, GPT-4 identified Ben Thompson as the author,” he added.
“It is difficult to identify the specific author of this blog post without more context or a link to the source. However, the content and style seem to resemble that of Ben Thompson, who is the author of the Stratechery blog, where he discusses technology and business strategy. The writing style, focus on historical context, and analysis of the banking industry could be indicative of Thompson’s work, but without more information, I cannot definitively confirm the authorship,” GPT-4 said.
And that wasn’t all. GPT-4 also correctly identified Bloomberg columnist Matt Levine with the first four paragraphs of an article authored by him.
And GPT-4 identified Seth Godin as the author when two of his short posts were shared with it.
These posts were all recent, so it’s not as though GPT-4 had previously seen the posts, and merely recalled who’d written them. On the other hand, GPT-4 had likely seen other examples of the writings of these authors, and based off the new four paragraphs of text, was able to predict who’d written them based on the writing style. It’s quite a remarkable feat, and only underscores how much insight and intelligence is contained in seemingly simple Large Language Models like GPT-4.