evanscahill.bsky.social
@evanscahill.bsky.social
Reposted by evanscahill.bsky.social
GPT3, which is to say OpenAI's mode two generations back which was barely good enough to generate text that didn't instantly clock as nonsense, was trained on *billions* of words.
This is like saying you're going to build your own aircraft carrier in your backyard.
This is like saying you're going to build your own aircraft carrier in your backyard.
January 9, 2026 at 3:41 PM
GPT3, which is to say OpenAI's mode two generations back which was barely good enough to generate text that didn't instantly clock as nonsense, was trained on *billions* of words.
This is like saying you're going to build your own aircraft carrier in your backyard.
This is like saying you're going to build your own aircraft carrier in your backyard.
Reposted by evanscahill.bsky.social
People keep insisting you can do this and you can't. If you train a model on only, say, a couple million words (which is basically the sum total of writing in Larian's games) you would not get output that is anywhere near coherent enough.
January 9, 2026 at 3:41 PM
People keep insisting you can do this and you can't. If you train a model on only, say, a couple million words (which is basically the sum total of writing in Larian's games) you would not get output that is anywhere near coherent enough.
Reposted by evanscahill.bsky.social
i will be screaming this until the moment i die.
Or more Marxistly, AI critique reifies a political-economic movement as a technological force. This mystification helps naturalize the disruption of existing labor arrangements while directing critique away from the movement seeking to remake social relations through AI.
November 3, 2025 at 8:11 PM
i will be screaming this until the moment i die.