Artificial Stupidity — Text CorpusThe training data for AS-F
and the tiny models.
One full collection run: 116.9 MB, 4,023,624 lines, 64 minutes.
Source
Collected
Lines
What it teaches
Twitch chat (live scrape)
36.0 MB
1,389,682
Thousands of people reacting in five words or less
Reddit comments
24.0 MB
419,077
Real arguments between real humans
YouTube transcripts
17.9 MB
548,858
24 channels of people talking nonstop
Twitch chat (HF dump)
15.0 MB
604,340
More… See the full description on the dataset page:
https://huggingface.co/datasets/ayushmaninbox/artificial-stupidity-corpus.
