top | item 40558244

(no title)

chewxy | 1 year ago

Most LLMs out there are very American in their writing mannerisms. The article even alludes to this:

> The AI did, however, try to sound like someone. It was folksy and upbeat, talky and pretend-excited

I've not seen an LLM, even when fine tuned that doesn't actually do that (Chinese LLMs excepted). There's something just inherently American about the instruction datasets that these LLMs are instructed with.

discuss

order

sentientslug|1 year ago

I’d wager it’s because most LLMs are trained on Internet data which can be very American-centric

barfbagginus|1 year ago

I got the AI to write excellent Romanian, both standard and rural dialects and styles. It wasn't hard. The standard remaining was a default, and I was able to dial in a rural character in less than 10 rounds of back and forth conversation about the area and background of the person and some of their speech mannerisms. These are broad strokes that the AI kind of filled in for me, not teaching it from scratch.

So just prompt it a little bit and spend maybe 300 to 800 words talking in the style you want it to emulate. And have an extended conversation where you nail in the style.

I bet you can do this to make it write in British, Australian, or Indian English styles. But I'm perhaps not really equipped to judge, since I don't know those styles natively.