Conversation

RE: https://mastodon.acm.org/@ACM/116929782064773116

The ACM wants to open up their library to LLMs for inclusion in synthetic text extruders. This is obviously massively misguided.

There's a form for you to tell them about your thoughts: https://docs.google.com/forms/d/e/1FAIpQLSeq7MFM7TlSXIaIXQnIltiQeWE_w3cl36HFPwql4S7u-drmzg/viewform

My response:
"Science is a deeply human endeavor based on collaboration and mutual respect. Giving machines based on massive appropriation of data _against many creator's explicit will_, machines that are deeply problematic in their tendency to hallucinate even more legitimacy by allowing them access to the ACM Library is massively misguided.

The ACM as scientific organization must care about proper attribution of work and transparency in where certain positions and data come from. That is in fundamental opposition to how LLMs work. If ACM takes their own foundational values (and itself seriously) accepting LLMs as valuable tool is out of the question.

Not only is that an insult to the many researchers who subscribe to the fundamental processes for collaborative science and therefore do not see LLMs as the benefit that OpenAI, Anthropic and a few others claim them to be. The legimatization of the deeply unethical systems that Amnesty International recently called "fundamentally incompatible with human rights" would be the ACM giving up its role as defender of quality in engineering and science but doing PR for a handful of unethical companies.

The supposed benefits are marginal and often potential at best while the harms are real and material. This is the wrong call by ACM and I urge Leadership to rethink it and focus back on its core values."