11 ms·

> _Almost_ brilliant enough to ignore the accent of Evgeny Morozov. Do you have any particular issue understanding his accent? Is it distracting? I know this c

> _Almost_ brilliant enough to ignore the accent of Evgeny Morozov. Do you have any particular issue understanding his accent? Is it distracting? I know this criticism and I've held it with respect to other podcasts (or for less: just bad microphones or engineering making listening grating). But... he's perfectly fluent in English, I have no trouble understanding him, and I don't find his accent distracting either. The engineering on the audio is good as well. Not sure what the problem is here.

breck 2y ago
Click "Watch the Trailer" Set it to 2x speed. Feed the audio to a voice-to-text engine. See the accuracy difference between when he is speaking vs his subjects.
- 1propionyl 2y ago
  Why would I do that? I don't care about how accurate a STT engine is on his voice? Seems like an artificial benchmark. All that teaches me is that my auditory system is still far superior to SotA in STT, which... isn't surprising? If anything, your benchmark seems more like an indictment of how such STT engines have been trained (bias in training data, etc).
  - breck 2y ago
    > Why would I do that? I told you that I found him hard to understand. You disagreed. I suggested one way to get an objective measurement. Maybe there's some web app that auto-grades this sort of thing? Like a Grammarly but for audio quality? I'm not sure. It's an interesting idea.
    - 1propionyl 2y ago
      That's not an objective measurement of how easy he is to understand for humans. That's a predictive score from a model trained on some set of audio recordings of how likely it would be for it to have produced that audio. This can only possibly be construed as a "measurement" with respect to the model. It's a measure of how the model relates to the input. I'd claim that's not a good proxy for human comprehension at all. In text, it works a good bit better as text data and grammar are far less rich than speech with intonation, accent, cadence, etc.

breck 2y ago

Click "Watch the Trailer" Set it to 2x speed. Feed the audio to a voice-to-text engine. See the accuracy difference between when he is speaking vs his subjects.

1propionyl 2y ago
Why would I do that? I don't care about how accurate a STT engine is on his voice? Seems like an artificial benchmark. All that teaches me is that my auditory system is still far superior to SotA in STT, which... isn't surprising? If anything, your benchmark seems more like an indictment of how such STT engines have been trained (bias in training data, etc).
- breck 2y ago
  > Why would I do that? I told you that I found him hard to understand. You disagreed. I suggested one way to get an objective measurement. Maybe there's some web app that auto-grades this sort of thing? Like a Grammarly but for audio quality? I'm not sure. It's an interesting idea.
  - 1propionyl 2y ago
    That's not an objective measurement of how easy he is to understand for humans. That's a predictive score from a model trained on some set of audio recordings of how likely it would be for it to have produced that audio. This can only possibly be construed as a "measurement" with respect to the model. It's a measure of how the model relates to the input. I'd claim that's not a good proxy for human comprehension at all. In text, it works a good bit better as text data and grammar are far less rich than speech with intonation, accent, cadence, etc.