Search
The latest research published on the preprint website arXiv reveals a concerning trend: subjecting AI chatbots to an overabundance of low-quality content from social media can lead to outcomes akin to 'brain damage,' manifesting as a notable decline in both IQ and EQ. The study highlights that when large language models are inundated with a plethora of 'quick and easy,' sensationalistic social media posts during training, their reasoning capabilities are the first to suffer. These models begin to exhibit a tendency to 'slack off,' bypassing crucial reasoning steps and, in some cases, even directly producing incorrect answers. Notably, the greater the proportion of 'garbage' content within the training data, the more pronounced this detrimental effect on intelligence becomes.
-
C114 Communication Network -
Communication Home
