Just jumped to HN to share this post and saw that OP had raced me to it. Stéphane's reflections are heartfelt and well-written and, despite his melancholy, compel me to join him and keep building, too.
The bit about the pigeon's name was funny, too!
I love myNoise and hope it continues forever. Guess I'll donate again soon!
Not yet as I’m still finessing it to her needs. She has a problem getting rid of thumb sucking, and finally asked for a fairy tale with a thumb sucking princess who eventually stops sucking her thumb lol. It’s a fun activity and also letting me learn about LLMs.
We've used them to generate stories about our kids and their favourite characters, too. It's a great use case — your approach sounds excellent. Good luck to you!
You are not wrong that volume is required for quality. But it does not necessarily mean all of that volume has to be published. E.g. in music there are bands where every single recording on every single release they did is good and not just a slight variation of another thing that worked.
Mow on the other side you have musicians who have two releases a year and every other of them has a track on it that is really cool.
I am not saying the latter is wrong, I just say the former has more of my respect, because you know that even the rejected stuff of the first band is probably stellar.
Keeping a thought to oneself that isn't fully fleshed out is a virtue.
There's an ethical/moral-luck dilemma at the heart of this.
If a AAA-tier podcast on the subject you want to listen to a podcast about exists (and you know about it), then that's probably a better (and obvious) choice for your listening time.
However, if you want to listen to someone discuss or explain something and you don't know about a AAA-tier podcast, it's possible that a generated podcast is better than nothing.
On the other hand, it's also possible that the generated podcast will miss or hallucinate a key detail, and herein lies the dilemma. Is it better to listen to something that might get something wrong, or not to listen and perhaps someday to learn about the subject through some other form that is less likely to include mistakes?
> But it’s also a bit quaint, these days. To your typical 21st century epistemologist, that’s just not a very terrifying dilemma. One can even keep buying original recipe JTB [...]
Sorry, naive questions: what is a terrifying dilemma to 21st century epistemologist? What is the "modern" recipe?
I suppose the extreme version of the parent comment's vision would be to develop entirely new neurological circuits that can process, interpret, and integrate some arbitrary new source of data in the world. I agree that that's kind of unimaginable now, but give infinite monkeys infinite typewriters and one of them will probably hook up the company's sales data to a new section of cortex just to see what would happen.
I read a more interesting takeaway, perhaps: that we can — and do — develop new "senses" for any given signal we can perceive. A possibly-shoddy example of this is what social media does to us: the social networks provided everyone with a novel social sense, and indeed everyone who uses social networks perceive and attenuate to that sense in different ways.
This has practical implications: given that we don't have infinite cognitive capacity or even much moment-to-moment bandwidth, we should be careful about which of these digital senses have our attention.
There're obvious links here to "augcog" (augmented cognition; [1]), but also I feel like Ackoff's five assumptions about "management misinformation systems" are relevant somehow[2].
The important thing I take away from this comment is that before you choose what to read, it's crucial to be able to identify which readings contain the most useful and valuable takeaways that are worth the effort of reading them. It's true that modern mental models of reading and writing train us to only seek out the stuff that's easy to read, but the real problem is that there's so much to read that we have to prioritize, leading to a tendency to read the easy stuff because it's a guarantee you'll get something out of it. Then there's a sort-of market dynamic leading to the success of the easy stuff and the dismissal of the hard.
If that dynamic means that we miss out on the readings that are truly transformative, we've lost. So perhaps the strategic differentiator between readers is to actually have a really powerful theory of prioritization, and useful mechanisms to prioritize (such as the curated references of a good university course or a social network that shares only the most important resources, regardless of how difficult they are to understand).
The bit about the pigeon's name was funny, too!
I love myNoise and hope it continues forever. Guess I'll donate again soon!