Thread by @OwainEvans_UK

1.Language models could become much better literary stylists soon. What does this mean for literature? A highly speculative thread.

2. Today models have limited access to sound pattern / rhythm but this doesn't seem hard to fix: change BPE, add phonetic annotations or multimodality (CLIP for sound), finetune with RL from human feedback. GPT-3 is a good stylist despite handicaps!
gwern.net/GPT-3#rhyming

3. There are already large efforts to make long-form generation more truthful and coherent (WebGPT/LaMDA/RETRO) which should carry over to fiction. RL finetuning specifically for literature will help a lot (see openai.com/blog/summarizi…, HHH, InstructGPT)

4. Language models will be better literary stylists than nearly all humans. Humans with good ideas (but merely decent prose skills) could use models to write great literature (from the perspective of today - 2022)

5. Thus our perspective on literature will change -- like change in painting after photography. Maybe a shift to generalized autofiction/live-tweet/streaming, where human author writes but also shares life in other modalities (that AI can't emulate).

6. Maybe a shift to an intense literature of ideas. Today there's a big audience for ideas if form is accessible and inspiring (great fiction) but not if form is dense, dry, impersonal, jargon-laden academic literatures.

7. Assuming humans are better than models at creating and collating ideas, then there could be a new explosion in the literature of ideas (where humans work with models to create compelling literary exploration of ideas). HT @peligrietzer for discussion.

Some inspiration for this thread: existing autofiction (e.g. Knausgaard and esp. last book of My Struggle), literature of ideas (Chiang, Stephenson, Stoppard, HPMOR), idea-heavy non-fiction with more accessible form (GEB, Selfish Gene, Marginal Revolution, ACX).

@threadreaderapp unroll