Why Critical Fallibilism? [Podcast]

Original link to watch on Descript’s website

Unscripted. I just sat down and talked for a while. All I planned was to talk about the question in the title.

I didn’t put it on YT yet because of experimental features. Please give feedback on any issues with the audio, or let me know if it seems fine to you. The biggest concern is due to automatically removing filler words.

I made the audiogram (the visualization to go with the audio) in Descript (you just pick from a few templates). I used the shorten silences feature (to 3s, which is apparently the maximum) and the remove filler words feature (like “um”, “uhh”, “so”, “you know”, “I mean”, “kind of”). Without much review, I found a place where Descript caused a problem: I had around a 2 second gap between sentences and started the second sentence with a filler word. When removing the filler word, Descript got the gap between sentences down to nearly zero so the words run together and it’s hard to hear. I also used Descript’s feature for removing background noise. I want feedback on whether these features broke anything.

The transcript is auto-generated by Descript. I fixed a couple words near the start (in particular, changing “critical fellows” -> “critical fallibilism says”) but there are plenty more transcription errors throughout. That is OK. I don’t need reports of transcription errors. You can just give a general impression: are they are bad enough that you’d rather not see the text at all? Is the visible transcript worthwhile? Personally I often watch YouTube with auto-generated subtitles enabled. My first impression is YT’s transcription might be better than Descript’s.

Descript’s audiograms have a 30 minute limit for some reason, so you can’t use them long podcasts.

I wrote a few notes on screen while talking. The audiogram doesn’t include the screencast so I pasted my notes below (there really wasn’t much to see, so I don’t mind leaving the screencast out).

  1. It builds on CR, Oism, TOC.
  2. Binary approach. Rejects degree arguments/evaluations.
  3. The “problem of induction” is over 2,000 years old and remains unsolved. Binary approach offers a new way to deal with it which isn’t already known to fail.

Many alternatives are bad because they focus on psychology without any improvements to the underlying philosophy parts. They focus on teaching rationality better, and getting people to actually do it (be motivated, form habits, etc.), but if their concept of rationality is flawed then they’re solving the wrong problem.

I didn’t like how short the transcription snippets were on Descript’s video – like less than 8 words on-screen most of the time, sometimes only 4-5.
Compared to YT subtitles, it makes it harder to listen to at 1.5x+. The problem I think is that YT subs will stay on the screen longer (and the bottom line animates up), so a word never disappears while you’re reading it. But with the Descript transcript, that happened plenty and meant I had to concentrate more to follow along – usually when there was a string of 10+ shorter words spoken quickly.

In that case, which is mb 10%-20% of the time, I think having descript’s transcript in-video makes it harder to read/listen at the the same time. The other 80% of the time, the hardcoded subs aren’t that helpful though because I found your speech was usually slower in those bits.

WRT rather not having them there at all: I can just look away, so i didn’t mind it (I did this to test if it was easier to harder to listen with the subs).

I don’t mind Descripts video-transcript in general (like visually, appearance-wise). I’d like the font size to be smaller (which would probs be better for a YT vid anyway). But if the podcast were on YT I’d just use autogen subtitles or not pay close attention to the visual part. IDK if having 2 sets of subs would be distracting. IMO Descript’s video is better than like just a black screen, but mb that’s just taste. Having only the audio-waveform visualization without the words would work too.

Also, YT subs tend not to have punctuation, I think, but descript adds lots of punctuation, like periods, which is a bit distracting when it clashes with what you’re actually saying.

I listened to the first 1/3 or so before writing this.

I don’t really understand this because the screen space isn’t needed for anything else. Did you full screen it? It’s really big that way but can’t you just use the small player if you’re sitting close to a large screen? And if you’re on a small screen, e.g. a phone, then the large font size is better.

The other 3 descript templates have 3 or 4 lines of text at once, or 1 word at once. There’s no option without words.

YT subtitles don’t animate up for me with Mac Firefox. What are you using?

I think I’d prefer non-animated anyway. It’s more like RSVP. If you start reading text when it first appears then you should have enough time to finish. If you don’t, then I think animation won’t help – the rate of words is just too fast for you (assuming the speaker is talking at a constant rate instead of e.g. saying quick points with breaks in between).

Ah that’s interesting, yeah. Descript also guesses where paragraph breaks are (it’s visible for the transcript in their editor).

If I put it on YT, you’ll be able to enable YT subs on top of the Descript audiogram. Two sets of subs on screen might be weird/annoying though.

Notice any audio issues like words too close together or anything that seems both bad and probably not what the raw audio was like?

My preference for a smaller font is just so more text can fit on the screen. No point though if it were just a smaller font so that it took up less space (but had the same number of words showing at any one point).

I mentioned font size specifically b/c maybe that’s a parameter you can tweak – I can see why that was unclear on my part.

FF on Windows + Chrome on a different Windows box.
I have Video Speed Controller installed on both (though I don’t think that should matter).

Here’s a 7s clip of what it looks like:

Maybe I’m just used to this [animated subs] and wouldn’t mind so much otherwise.

Nope – didn’t notice any cuts and the occasional ‘long’ 2-3s pauses sounded good/natural. I didn’t notice any bits where it seemed obvious a filler word had been cut out either.
I listened to some at 1x and some at 1.5x (no noticeable audio issues either way).

Can you publish a new version with only 1 word on-screen at a time? I guess that the words would be in the same place like RSVP – that might solve the problem I had with the hardcoded subs in OP. (I’ve tried RSVP programs before and liked them)

Also, it makes more sense (IMO) to have YT subs + RSVP subs than very similar hardcoded/YT subs.

Edit: the animating YT subs work fine at higher speeds btw – the video above is at 1x.

1 Like

I didn’t hear anything wrong with the audio.

1 Like

I didn’t, but was also not paying super close attention. It seems fine to me.

I think the text was helpful; the transcription errors weren’t enough to eliminate its value.

I like big fonts but agree more of the text would also be helpful. Particularly, more lines of text and older stuff scrolling off rather than being replaced all at once.

1 Like

I posted it on YouTube and edited the YT link into the original post.

I like to listen to podcasts in the background so this is perfect for me. I didn’t find any issues with the audio.

1 Like

I think that Popper thought that he solved the problem of induction and I think DD also thinks that Popper solved it. Do you disagree? Or your solution improves on the incomplete solution that Popper gave?

Many alternatives to what? Critical falliblism or solution to the problem of induction?

I meant that making induction work is unsolved. Popper found a solution to epistemology that doesn’t use induction, rather than a direct solution to the problem of induction.