Post #4163457
2026-07-28 12:02 UTC
Replies (2)
-
@theodoraward@mastodon.social 2026-07-28 12:17
@zzt@mas.to it's a bummer because "enshittification" really did change my life. i don't really know how he gets from "copyright law is flawed and creators probably shouldn't rely on it to save us from corpohell" to "training LLMs without the consent of those who authored the training data is normal and good." it's a kind of legalistic literalism that outweighs ethical thought? i don't know why people need to redeem this technology. a couple typos aren't worth running public interference for this shit
-
@zzt@mas.to 2026-07-28 12:08
it’s also a bit wild to start a thread by praising chatgpt and then immediately go into a rant about local models, but that’s how cory works in a nutshell. he’s being real obvious this time, but that’s a symptom of cognitive surrender. pay close attention to the mechanism now that it’s very visible. he praises a proprietary model on frivolous grounds and then launches into a defense of his own use of local LLMs, also on frivolous grounds. who benefits from this framing? if you take doctorow seriously and try a local LLM, you will find it to be disappointing because local LLMs are notoriously terrible. but he kept you distracted enough that now the idea that chatgpt can do groundbreaking work in math (it can’t, it’s been proven again and again that it can’t, it doesn’t even make sense that it could) stuck around. so the local LLM didn’t work, better try the harder stuff!