Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

There is an undercurrent of skepticism and contrarianism bordering on anti-intellectualism on this forum - especially with those accounts started after the pandemic.

This is just one of dozens of studies that have shown that good cardiovascular health (a side effect of "vigorous" exercise) has a downstream positive effect on health.

And I personally can attest to that - the days I don't vigorously exercise (run, swim, BJJ, MT) are also the days I see the biggest cognitive declines - and I'm fairly young.



This should be quantifiable. Use an LLM to help identify this undercurrent you’re referring to in maybe 10,000 conversation threads using the API from the footer. A thread being a reply to a topic, and a singular chain of replies.

Then use that to train an RNN for sentiment analysis based on the values you’ve just described.

Then run the entire HN corpus through your RNN and chart the values over time.

For bonus points, incorporate the score of replies in the weighting of values.


You don't need an LLM for sentiment analysis.

I also don't want to pay for scraping and data ingest costs out of my personal pocket.

Feel free to do it if you want - it would make a nice blogpost.


I wasn't suggesting you need the LLM to perform sentiment analysis. The problem the LLM overcomes is generating the training data you need without having to trudge through massive amounts of material to mark a comment / thread with a 1 or a 0.

There's no scraping cost, the API is free. Your message, for example, is here: https://hacker-news.firebaseio.com/v0/item/40801233.json?pri...

Processing 10,000 threads is probably not more than a few dollars on OpenAI's GPT API.

The reason I don't do it is I don't see the trend you see.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: