Blog written by artificial intelligence takes top spot on Hacker News

An American student has reached a top spot on Hacker News with a blog written by artificial intelligence. Readers of that platform did not notice that the text was not written by a human. The student used the OpenAI GPT-3 language model.

His blog with posts written by GPT-3 has generated a total of 26,000 visitors, Liam Porr writes. The American student created three accounts on Hacker News to submit links to his own blog and give a few up-votes. One of the blog posts gained traction and made it to the top of the list. Under that post, entitled ‘Feeling unproductive? Maybe you should stop overthinking ‘sparked a substantive discussion from Hacker News users.

The vast majority of users did not realize that it was a text that was not written by a human being. There is someone who remarks that it appears to be a text generated by GPT-3, but that person is contradicted by others.

Porr explains to MIT Technology Review that he himself wrote a title and an intro text and added a photo. He let GPT-3 do the rest of the work and finish the story. Incidentally, this OpenAI language model is not freely available. The organization fears that it will be abused. That is why the beta version is only accessible to researchers and the use is examined. Porr was able to make use of it through a student who is doing PhD research and who has access that way.

Although Porr let the artificial intelligence do most of the work, he did make minor adjustments to the lyrics. For example, the AI ​​attributed a quote from Bob Dylan to someone else and there were a number of spelling errors in the lyrics.

OpenAI presented GPT-2 early last year. That language model had 1.5 billion parameters and was trained on eight million web pages. Artificial intelligence can write coherent texts by predicting the next word each time. GPT-3 was released in beta in July this year; this successor works with 175 billion parameters, making it the most advanced language model.