ACM is now arguing that all papers should be fed to LLMs: https://cacm.acm.org/opinion/now-is-the-time-to-give-llms-access-to-the-acm-digital-library/.
I'm thinking about responding to them. Is this a good response?
Once upon a time, there was an eager undergraduate student. He was passionate about computational linguistics as well as machine learning. He thought about combining both. He had applied to become a graduate student at Universite de Montreal, one of the most renowned research facilities for machine learning. His prospective supervisors were Guy Lapalme and Bengio Yoshua, a famous neural network scientist. Had everything gone according to plan, he would probably become a great AI scientist.
But this was not meant to be. He ended up with a PhD in human-computer interaction instead. Although his life is different from what he had believed, he occasionally thought about his alternate self who got to become a great AI scientist.
And then, he has to deal with many instances of cheating. His dream has turned into a nightmare. His long-forgotten past has come back to haunt him in a rather grotesque way. He is now facing many students who all claim that their nearly identical code is wholly original. They abuse AI, and in turn, AI abuses them back. The students can no longer recognize what real code looks like. It is also highly probable that many research papers that the students wrote were also AI-generated. But he could not be any wiser.
Peer reviews also became nightmarish. He had to wade through haunting papers that were seemingly AI-generated. While many great statisticians were trying to settle the Bayesian v. NHST debate, little did they know that the cutting-edge statistical analyses are now just p-values without any statistics and descriptive statistics are seemingly sprinkled randomly and arbitrarily. The reviews from other scientists were also odd, and occasionally nearly identical.
This almost-AI-scientist also has made a terrifying discovery while working through HoloLens sensor data. From a haunted AR-enhanced museum, he found data sets that repeatedly laugh at the combined human-assisted AI and AI-assisted human attempts to parse them. As it turns out, the data require some extreme out-of-the-box thinking. AI and humans must work together. However, the "AI" here is not the LLM-based one. Rather, it is an aggressive form of exploratory data analysis (EDA).
What the ACM is doing here is not what I have been doing. Instead, it is pushing all work onto AI. This is an asymmetrical relationship where the AI is the Giving Tree being cut down to its stump. This is not respectful to AI nor its pioneers.
He does wonder about his alternate self sometimes. Perhaps, his alternate self is sleeping easy because he is making a lot of money and getting all the accolades. Or perhaps, he is sad knowing that his work is being used to degrade education, peer review processes, and more.
This scientist with a low citation cannot dictate what the ACM will do next. Afterall, he let his ACM membership lapse. However, he is making a humble request: Please leave his ACM papers alone.
Also, you guys should totally check out Neurosama streams - especially the cooking shows. You may find the experience enlightening in some ways.
blakekathryn
baratheonpilled
izzy806