UC Berkeley student Liam Porr used the GPT-3 algorithm to write blog articles on the web-based aggregator Hacker News.
Porr did the title and intro, add a photo, and GPT-3 did the rest. For two weeks the blog was visited by over 26 thousand users, 60 people subscribed to the author, the student said.
“People say that GPT-3 is often incoherent and illogical. But that doesn’t stop them from reading and liking the texts, ”says Porr. The first text, made by GPT-3, topped the list of the most read on the site and collected 47 likes (at the time of writing this note 198 — approx. Vc.ru), he points out.
Only one user contacted Porr and specified who the author of the text was. Some pointed out in the comments that the text was written by GPT-3, but they were minus and even accused of incorrectness, says the author of the blog.
“It looks like GPT-3 or some other algorithm wrote it. Text about nothing […] — “You may be new here, but your rude comment is not acceptable in this community. If you disagree with the text, be polite and provide arguments, and do not offend the author “
2. Fundamentals of AI, ML and Deep Learning for Product Managers
3. Roadmap to Data Science
4. Work on Artificial Intelligence Projects
According to Porr, he was a little uncomfortable with the GPT-3 writing the text, but also funny. He wanted to show the readers the possibilities of the algorithm, realizing that they would not take their word for it.
“People need some kind of proof before they start taking things seriously,” he says.
The GPT-3 algorithm was created by OpenAI, an American laboratory for the study of AI. GPT-3 is trained on 570 GB of text information and can write texts in English.