What a cool project! I listen to the vast majority of my reading these days and am perpetually out of good things to read.
The linked audio is reasonably high quality, and more importantly, it doesn’t have some of the formatting artifacts that other TTS programs have. Well done.
Your story for why this is a potentially high impact project is plausible to me, especially given how much you’ve automated. I have independently been thinking about building something similar, but with a very different story for why it could be worth my time to do it. That means this could be a different story for why your thing was good :), which I thought I’d share
My story was that the top-performing people in a given cause area are large fraction of the valuable work, if you buy power law type arguments. By definition, their time is a lot more valuable than average. But it is also more valuable for them to be better informed, because the changes they make to their decisions by being better informed are leveraged by their high work output or its consequences.
If you buy this story, I think you wind up focusing on figuring out how to cater what is audio-fied to what would be useful to the most productive people in EA. So like, what do top AI safety researchers wish they had time to listen to. I’d bet that this is actually a very different set of things than Forum/ LW posts.
When I started to do my thing, I suspected that a lot of the researchers who are doing the best work would benefit from being able to hear more academic papers, from arxiv for example. But IMO the key problem is that these don’t get read well because of formatting issues. I think this is a solvable problem, and have a few leads, but it was too annoying for me to do as a side project. DM me if you’re interested in chatting about that
Side point: this view of why this is high impact also speaks to letting the top people in question choose what they listen to, which looks more like an app that does TTS on demand than a podcast feed. This happens to avoid copyright issues, if the existence of other TTS apps is any indication.
You might be able to hack together an equivalent solution (on both copyright and customization) without needing to develop your own app by having a simple website that lets people log in and makes them a private RSS feed (compatible with most podcast players I think, though not confident in any of this). Then if they input a link on the website its compiled and added to their RSS feed for use in the player. If you had an api for calling your TTS script (and had solved these formatting issues) I or someone else could probably hack something like this website together pretty fast
(Sorry, when I said your story for impact was “plausible”, in my head I was comparing it to my own idea for why this would be good, and I meant that it was plausibly better than my story. I actually buy your pitch as written, seems like a solidly good thing; apologies)
What a cool project! I listen to the vast majority of my reading these days and am perpetually out of good things to read.
The linked audio is reasonably high quality, and more importantly, it doesn’t have some of the formatting artifacts that other TTS programs have. Well done.
Your story for why this is a potentially high impact project is plausible to me, especially given how much you’ve automated. I have independently been thinking about building something similar, but with a very different story for why it could be worth my time to do it. That means this could be a different story for why your thing was good :), which I thought I’d share
My story was that the top-performing people in a given cause area are large fraction of the valuable work, if you buy power law type arguments. By definition, their time is a lot more valuable than average. But it is also more valuable for them to be better informed, because the changes they make to their decisions by being better informed are leveraged by their high work output or its consequences.
If you buy this story, I think you wind up focusing on figuring out how to cater what is audio-fied to what would be useful to the most productive people in EA. So like, what do top AI safety researchers wish they had time to listen to. I’d bet that this is actually a very different set of things than Forum/ LW posts.
When I started to do my thing, I suspected that a lot of the researchers who are doing the best work would benefit from being able to hear more academic papers, from arxiv for example. But IMO the key problem is that these don’t get read well because of formatting issues. I think this is a solvable problem, and have a few leads, but it was too annoying for me to do as a side project. DM me if you’re interested in chatting about that
Side point: this view of why this is high impact also speaks to letting the top people in question choose what they listen to, which looks more like an app that does TTS on demand than a podcast feed. This happens to avoid copyright issues, if the existence of other TTS apps is any indication.
You might be able to hack together an equivalent solution (on both copyright and customization) without needing to develop your own app by having a simple website that lets people log in and makes them a private RSS feed (compatible with most podcast players I think, though not confident in any of this). Then if they input a link on the website its compiled and added to their RSS feed for use in the player. If you had an api for calling your TTS script (and had solved these formatting issues) I or someone else could probably hack something like this website together pretty fast
(Sorry, when I said your story for impact was “plausible”, in my head I was comparing it to my own idea for why this would be good, and I meant that it was plausibly better than my story. I actually buy your pitch as written, seems like a solidly good thing; apologies)