My impression is that some people were put off by the length of the articles in Late 2021 MIRI Conversations. Personally, I've used my iPhone's text-to-speech functionality to listen to these and similarly long LessWrong posts as I do other things. After someone else commented on how convenient that seemed, I thought I should try posting a text-to-speech audio version of "Ngo and Yudkowsky on alignment difficulty" and see if that made the content more accessible. 

If you find TTS audio versions of longer posts helpful or have other feedback, please let me know. I'm planning to generate TTS versions of the other MIRI conversations after getting feedback here. In the future, we may even want some sort of integrated TTS service for long LessWrong posts. Edit: thanks to Steven Byrnes for pointing out that we already have such a service from the Nonlinear Library. Here's their version of "Ngo and Yudkowsky on alignment difficulty".

Here is a SoundCloud link for my version.

The mp3 files are available at this Google Drive folder.

I generated the audio files with Amazon Polly using the neural version of the English/US voice Joanna. 

Following TTS audio of technical discussions is difficult at first. I've used my iPhone's TTS for years, and it still took me a few minutes to adapt to the Amazon voice. I suggest listening for at least 10 minutes, and not getting too invested in following all the details, especially at first.

I've striped out the timestamps on the posts, since they're difficult to follow and distracting in an audio-only format. If any of the participants would like me to add them back, make other minor changes, or remove this post entirely, I'd be happy to oblige.

3 comments, sorted by Click to highlight new comments since: Today at 8:55 AM
New Comment

Cool! Just curious: Is there something wrong with the Nonlinear Library version, or had you not heard of Nonlinear Library, or did Nonlinear Library not do those posts?

To be fair, there was indeed something wrong with our version! It was so long it messed up our system and we've only now fixed it and it's released in three parts.  Along with the other Eliezer, Richard, and Paul conversations 

I'd not known of the Nonlinear Library. Thank you for letting me know about it!

The Nonlinear Library mostly fills the gap I was aiming to address with this post and is more or less what I was suggesting with "In the future, we may even want some sort of integrated TTS service for long LessWrong posts."

I do think the Nonlinear Library versions could benefit from some more pre-processing. E.g., removing the exact timestamps of the discussion posts. Also maybe adding descriptions for figures/images in the original texts.