Computer scientist Geoff Hinton challenges the notion that chatbots are just glorified autocomplete, arguing that to predict the next word accurately, chatbots must actually understand the context. Hinton’s perspective made the author reflect on their own work, which sometimes involves offering shallow, associative responses akin to chatbot autocomplete. They find this approach useful, as it can lead to unexpected insights or help represent problems not previously considered. However, the author recognizes that there are moments when deeper, logical thinking is required. They emphasize that associative reasoning and logical understanding are distinct and should not be conflated.
https://statmodeling.stat.columbia.edu/2023/11/18/i-disagree-with-geoff-hinton-regarding-glorified-autocomplete/