/
lectern.tv
Play15s30s1x0:00/--:--
DoneWatched
Expand
University College London·Stevan Harnad
1:10:24
0:00/1:10:24
Type below to lock the video timestamp, or press S to capture a slide.
Understanding Understanding and Why LLMs Don't
Stevan Harnad critically analyzes the remarkable capabilities of large language models while arguing that their performance does not equate to genuine understanding or meaning-making.
Large Language Models (LLMs) have remarkable and completely unexpected capacities: they can chat with human beings about just about anything, almost indistinguishably from another human being – and about more things than any human being can or ever could chat about. We know how LLMs do it [next-token training on an enormous database of human text] but it is still astonishing that they can.