Content
@
0 reply
0 recast
0 reaction
Dan Romero
@dwr.eth
Has anyone successfully "trained" (?) an LLM on an author's writing style? e.g. take all blog posts, interviews (transcribe with Whisper) and tweets from Paul Graham and turn into an LLM?
9 replies
1 recast
9 reactions
woj — q/dau
@woj.eth
https://x.com/daraladje https://x.com/zebulgar/status/1712487141484450235
1 reply
0 recast
4 reactions
jacob
@jacob
new dwr blog posts are about to be next level
0 reply
0 recast
3 reactions
Harry
@htormey
Greg has been making some good content around this theme: https://x.com/GregKamradt/status/1720510376889188731?s=20
0 reply
0 recast
3 reactions
Alok Vasudev
@alok
that’s basically what happened when all the YC partners and founders started blogging trying to mimic pg’s voice
0 reply
0 recast
3 reactions
Alessandro
@azeni
Would more like fine-tuning an existing model. I suspect with the new GPT builder it'll be fairly easy to do this for any of your favorite authors.
0 reply
0 recast
2 reactions
Steve
@sdv.eth
Depending on your definition of success, in 2020 someone trained an LLM using GPT-2, ~30k tweets, subreddit posts (mostly rants) of a controversial esports player. The outputs are hilariously outlandish yet very characteristic of the "author." https://twitter.com/deepleffen
0 reply
0 recast
2 reactions
not parzival
@shoni.eth
only for tpot tweets which are outdated now- fine tuned And I think gpt especially with the new token limit would do a pretty good job in a chat session alone
0 reply
0 recast
1 reaction
robin
@robin
played nanoGPT which allows you to get going with a training dataset of shakespeare's works. long term goal is to funnel in a group chat i have with some friends for a fun "what would X person say" microapp repo: https://github.com/karpathy/nanoGPT karpathy's tutorial: https://www.youtube.com/watch?v=kCc8FmEb1nY
0 reply
1 recast
7 reactions
Joshua Miller
@joshmiller
Here's an interactive advice column trained on (a subset of) an author (Chris Onstad)'s writing style: https://raybot.help
0 reply
0 recast
0 reaction
