skip to Main Content

OpenAI’s Language Model With Better Manners

GPT-3, OpenAI’s flagship language model, has an updated version called InstructGPT that is far better than its predecessor. This new version is not only better at following instructions, it is also less toxic overall.

With this better alignment, which means it is better at following instructions, the model throws out much less profanities and offensive language, wrong information and other issues. Language models learn from large amounts of text from the internet, but the problem is that much of the data they feed on is not the best reflection of humans on the internet.

Lead transformational growth by developing an understanding of exponential and digital technologies. Click to know more about the MIT PE Technology Leadership Program (TLP).

Back To Top