AI expert Gary Marcus has criticized the recent release of GPT-5, deeming it 'underwhelming.' Marcus argues that the model's incremental improvements fell short of expectations, suggesting that simply scaling up existing models isn't a viable path to achieving Artificial General Intelligence (AGI). He highlights the release and related research as indicative of a challenging week for the generative AI field, further reinforcing the idea that a new approach is needed to reach true AGI.