Sidechat icon
Join communities on Sidechat Download
In any other case, if someone said “I’ve created a piece of software that’s grossly inefficient. But it’s fine, just throw more resources/compute at it.” they’d be laughed at. Why are we willing to sacrifice the standards of software engineering for LLMs?
upvote 6 downvote

default user profile icon
Anonymous 1w

Like companies will boast that they’ve built billion dollar data farms & super computers to power their … chatbots. That feels like more of a concession than a flex.

upvote 2 downvote
default user profile icon
Anonymous 5d

It’s because they don’t want to look like they’ve fallen behind, and for most of these companies that’s the cheapest way to make an advancement. That said Google is doing it efficiently by making their own hardware for the LLMs, and DeepSeek managed to build their models with chips that are weaker than American ones. The innovations have already been starting to be made, it’s a matter of when the other companies will catch on.

upvote 1 downvote