> If Chess could be played perfectly, no one would find it interesting. Rather, our inability to use brute force—thus, the need to devise imperfect, efficient strategies—is what makes the game artful.
Are you familiar with Tremendous Trifles?
> If you could play unerringly you would not play at all. The moment the game is perfect the game disappears.
> It would be easy enough to generate a billion or a trillion novels, but there is no way to get nuanced, useful feedback from humans on each one.
This is true, but nuanced useful feedback is the sort of thing that Large Language Models are starting to become capable of themselves. The tiny memory (~3000 words) of modern LLMs is a serious limit in light of the millions (more?) of words that comprise an author's literary identity, but orders of magnitude shrink quickly in computing. Consider more closely the possibility of self improvement via solo play. We'd need humans in the loop to periodically look at output and to tweak the professor-model so that its critiques of the novelist-model are better. It need not be monolothic, either; with inevitable efficiency gains, some day each individual can finetune a model to personal taste, or to precise need.
> ... linguistic deviations stigmatized not because they are evidence of inadequate thought (since they aren't) but because of an association with low social position
The imminent democratization of language skills is undersung. Even just the implications for legal representation are immense. I'll be adopting your framing.
> the next wave of spammers and bullshitters
Between high quality personalized content and LLM's ability to analyze spam and bullshit, people can be exposed to less of it and be better equipped to defend against it. Maybe with the same optimism I would've said the same thing about the internet at large three decades ago and been disappointed today. The internet evidently contains deep wells of good clear knowledge, but folks tend to splash about and sip from the mudpuddles.
This horse appreciates the fresh water. Thank you for your providence.
> If Chess could be played perfectly, no one would find it interesting. Rather, our inability to use brute force—thus, the need to devise imperfect, efficient strategies—is what makes the game artful.
Are you familiar with Tremendous Trifles?
> If you could play unerringly you would not play at all. The moment the game is perfect the game disappears.
-GK Chesterton 1909 https://www.gutenberg.org/files/8092/8092-h/8092-h.htm#link2H_4_0005
> It would be easy enough to generate a billion or a trillion novels, but there is no way to get nuanced, useful feedback from humans on each one.
This is true, but nuanced useful feedback is the sort of thing that Large Language Models are starting to become capable of themselves. The tiny memory (~3000 words) of modern LLMs is a serious limit in light of the millions (more?) of words that comprise an author's literary identity, but orders of magnitude shrink quickly in computing. Consider more closely the possibility of self improvement via solo play. We'd need humans in the loop to periodically look at output and to tweak the professor-model so that its critiques of the novelist-model are better. It need not be monolothic, either; with inevitable efficiency gains, some day each individual can finetune a model to personal taste, or to precise need.
> ... linguistic deviations stigmatized not because they are evidence of inadequate thought (since they aren't) but because of an association with low social position
The imminent democratization of language skills is undersung. Even just the implications for legal representation are immense. I'll be adopting your framing.
> the next wave of spammers and bullshitters
Between high quality personalized content and LLM's ability to analyze spam and bullshit, people can be exposed to less of it and be better equipped to defend against it. Maybe with the same optimism I would've said the same thing about the internet at large three decades ago and been disappointed today. The internet evidently contains deep wells of good clear knowledge, but folks tend to splash about and sip from the mudpuddles.
This horse appreciates the fresh water. Thank you for your providence.