2025-02-22 00:57:23
Re @Dorialexander a few months ago (one?) training an LLM was almost always the wrong choice, unless that was the company’s core purpose. but with reasoning models, the calculus changes a bit, and i’m honestly not sure how, but i think training is the right strategy most of the time now
2025-02-22 00:54:55
Re @h_thoreson are they they ones that hired smokey the bear?
2025-02-19 22:51:09
Re @francoisfleuret more broadly, diversity matters. we’ve always believed that, but LLM research is showing that diversity is a key differentiator in dataset quality.
so why the drift into STEM-only education? that can’t be good
2025-02-19 18:51:09
Re @iamkylebalmer @kimmonismus should’ve called it DeepThoughts By Anthropic
2025-02-19 10:17:25
Re @h_thoreson it’s hard to imagine an administration getting so many things wrong so quickly
2025-02-19 09:24:54
Re @iScienceLuvr @Enscion25 i mean, diminishing returns is kinda baked into the scaling laws..
