At the heart of today's artificial-intelligence models are vast bodies of training data ' text, videos and images created by real people and used to teach models how to recognize patterns and generate content. People are certainly training AI systems ' but are they training us, as well' A growing number of papers report that people tend to pick up writing patterns, reasoning methods and even opinions from the large language models (LLMs) they use. Some researchers say that this influence threatens to create a sameness among human writing and warn that the effect could even extend to text written by humans who aren't first-hand AI users. 'If people around you are interacting with these LLMs and adopting their writing styles, perspectives and reasoning, at some point it would surround you so much that it would seem like the more socially correct way to frame information,' says Zhivar Sourati, a computer scientist at the University of Southern California in Los Angeles and co-author of...
learn more
