RSS Meets GPT-3

https://piqoni.bearblog.dev/rss-meets-gpt-3/

by lexoj • 3 years ago

3 3 3 years ago

piqoni.bearblog.dev

zhte4152• 3 years ago

Point 4 in the example summaries is quite relevant for this function:

Is learning 'knowing' what an article said having read a summary?

If a reader decides it is not [and this is a point aside from whether the summary's accurate], or not sufficient, and desires to read more, what more information has been provided by a GPT-summary that could not be provided by a headline in the first place?

If it's not provided in the headline, are headline writers missing a trick so much so that GPT-summaries are needed?

If a headline's not sufficient space, why can't articles provide an abstract/summary that makes this redundant?

I feel these are increasingly pressing questions for writers and publishers that don't want to be summarised out of content and context.

lexoj• 3 years ago

Apart from not being "fun", I believe writers don't write summaries because they don't want you to skip their article. It's the same as "this book could have been a blog-post" issue.

Regarding headlines, sometimes I think they are intentionally cryptic/clickbait-y because that's how authors intrigue us and make us click them. So a summary it's useful in those cases, unfortunately.

But there are other kind of titles which benefit from summaries. Say, "Postgres 16" is released. In that case, I'd love to read three sentences describing the highlights instead of clicking it and reading the detailed release log.

zhte4152• 3 years ago

Absolutely, for both cases.

I'm reminded of RSS back-in-the-day, 2007ish a bit before, when a good proportion of everything had RSS with full text summaries. Then publishers, both large and even smaller blogs, decided they'd prefer to give their readers a worse experience and moved to publishing only teasers, often that 'read more' stuff, or even just headlines, in their RSS feeds. Because they didn't want their readers reading it not on their property, even if that property was only a hit counter.

I can expect a similar backlash to GPT summaries.

The Postgres people may care less than the long form article publisher. I feel that the momentum will be towards better summaries, and these need not be any less intriguing - if an academic paper can manage a meaningful summary in a paragraph or two so can a major publisher - there will be some that don't adapt (and some, that due to a readership, won't have to) that will kick and scream as their boat takes on water.

RSS Meets GPT-3

Related Stories