Mastodon Feed
Boosted by glyph ("Glyph"):
amethyst@n7.gg ("Amethyst 🌸") wrote:
Reviewing conference talk proposals is wild in the age of LLMs, and exhausting in ways that it never was in the past. Never knowing if what you read was directly from the mind of a human, “touched up” by an LLM, or fabricated whole cloth by a stochastic lying machine.
Low effort proposals in the past were frequently just a couple sentences and nothing else, but now “low effort” can also produce hundreds or thousands of words for reviewers to sift through.
Proposals *about* LLMs take it to the next level. A human (most likely) wouldn’t simultaneously claim that an LLM both “generates boilerplate” and “reduces boilerplate”, but an LLM generated proposal about LLMs will.