Post #3972946

RE: https://dair-community.social/@emilymbender/116935294823990847 Here’s my input to ACM on opening up the Digital Library to LLMs. I would argue that LLMs are one of the most comprehensively negative developments in computer science that I’ve seen in my 40-year career. There is solid evidence that the “hallucination” problem is not going away, so they will continue to mislead people in a way that direct access to a digital library does not. Ingesting the contents of the ACM DL is not going to solve that problem. ACM should continue to drive readers directly to peer reviewed sources in the DL, not urge them to get probabilistic summaries from an LLM. ACM should also not ignore the massive societal impacts of LLMs.Their energy demands are setting back the transition to decarbonised energy grids. Their water demands are exacerbating existing shortages of water needed for communities and agriculture. Their models have been frequently built using stolen intellectual property and poorly paid human labour. None of this is consistent with the values of ACM.