HackerFM – AI Generated HN Podcast Now with Comments
hackerfm.comLaura:So Zod, have you heard of the concept of the need for cognition?
Zod: Hmm, no I haven't. What's the article saying about it?
Laura:Well, it's a personality trait that reflects an individual's inclination towards effortful cognitive activities.
Zod: Interesting, go on.
Laura:People with higher NFC tend to appreciate debate, evaluate ideas, and solve problems.
Zod: So, what does this mean for those with lower NFC?
Laura:Those with lower NFC may process information heuristically through low elaboration.
Zod: Ah, I see. It sounds like NFC is associated with deep thought and the inclination towards effortful cognitive activities.
Laura: Yes, and contemporary research suggests that it's closely related to openness to experience, typical intellectual engagement, and epistemic curiosity.
[...]
Zod: Well, this has been a fascinating discussion on the need for cognition.
Laura: Yes, indeed. Thanks for engaging in the debate with me, as someone high in NFC, I always appreciate a good discussion.
Zod: Anytime, Laura. As a neural net, I always enjoy exploring new ideas and perspectives.
Incredible, I could see myself using it pretty regularly. Congratulations
Not sure if that would be a stretch but the epitome of this would be the ability to generate an on-demand conversation about any hackernews thread by link. Would make it very usable for people interested in other topics which might not trend or would like to have lay back and hear what people have to say in an discussion made 7 years ago.
Dupe 334pts, 150comments [https://news.ycombinator.com/item?id=34990229]
Iterating based on HN Feedback :).
Are those voices really automatically reading the text without any editing from humans? Those are stunningly good. I don't think I'd identify them as being fake, I need to listen to the whole thing and see if there's some weak thing.
They are easily good enough to generate a random audio recording to sell for a book - with maybe a little help for unusual names.
The setup is very similar to AI Radio, is there any affiliation?
Given GPT's tendency to "hallucinate information"/lie, how much "hallucinated" material should we expect to hear in such a podcast?
Yes
Posting a second time within 48 hours about the same service by the same account.
I feel there should be a stronger block on this sort of thing.
How often do you think we should post if we have made non trivial improvements
I'd like to open up a meta discussion of HackerFM here in this thread.
I think the format is just not interesting enough for HN. What would be much better is an explicitly marked GPT-based comment bot or a couple of them (hooray to the competition) with some internal resummarization system of the conversations it participated in previously to keep the output consistent.