Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

NotebookLM's automatically generated podcasts are surprisingly effective

simonwillison.net

Post reply on HN
Print view View original

401–410 of 511 posts

Re: NotebookLM's automatically generated podcasts are surprisingly effective

#401

Post by AnIrishDuck »

gcanyon wrote: post #377
fooblaster wrote: post #347

Earlier quoted context omitted.

go is a game with an obvious score function which can be used to construct a loss, well defined moves, and total visibility of the board. It is less obvious how to write a score for creativity in art or music, nor does it have well defined bounds on what is considered a legitimate construct of either. Just because computing hardware lets you multiple matrices faster does not mean we have the means to solve all proble…

> go is a game with an obvious score function which can be used to construct a loss, well defined moves, and total visibility of the board This is literally the opposite of true, and the main reason go computers were getting destroyed by humans for almost twenty years after deep blue took down Kasparov.

Yes, technically. But the broader point is true. Go is a game with well-defined win and loss conditions that can be automatically evaluated.

This is critical for game-clock-eons of unsupervised self-play, which by most accounts is how AlphaGo (and other systems like AlphaZero) made the leap to superhuman levels of play.

But it is entirely different from subjective endeavors like writing, music, and art. How do you score one automatically generated composition vs another? Where is the loss function?

Re: NotebookLM's automatically generated podcasts are surprisingly effective

#402

Post by malignblade »

pessimizer wrote: post #370
huijzer wrote: post #132

Earlier quoted context omitted.

> The reason so much writing, podcasting, and music is vulnerable to AI disruption is that quality has already become secondary. I think that has always been the case, we just tend to compare today’s average stuff with the best stuff from earlier days. For example, most furniture pictures from the 60s and 70s are from upper middle class homes. If we listen music, we listen to Queen and not some local band from Alabam…

> I think that has always been the case, we just tend to compare today’s average stuff with the best stuff from earlier days. I agree with this of course, because generally nobody remembers the bad stuff unless it was the worst . I beg to differ with music, though, because there's an opposing effect: we tend to be left with the most marketed music, which was usually a cheap knockoff of something interesting going on…

You're raising a good point about how "best" is defined.

If you ask most people, they are by definition more likely to connect with broadly disseminated cheap knock offs than they are with whatever 'legit' inventive underground creator, simply because they've heard the former and not the latter.

Just a mental exercise: If you ask 1000 people if they prefer Knock Off or Original, and 900 say Knock Off, which one was better? If the answer is still Original, by what metric do we measure quality?

Re: NotebookLM's automatically generated podcasts are surprisingly effective

#403

Post by vietjtnguyen »

One of my favorite ChatGPT uses is voice chat during long drives as a pseudo, albeit interactive, podcast to learn about various technical topics at the edge of my knowledge base. This podcast generation is pretty amazing, but hopefully they make the "competency level" of the hosts tunable. One thing I love is being able to guide ChatGPT to the technical level I'm looking for. Maybe I'm just bad at finding podcasts, but only Signals and Threads [1] really has that interesting depth.

[1]: https://signalsandthreads.com/

Re: NotebookLM's automatically generated podcasts are surprisingly effective

#404

Post by drusepth »

I hate podcasts because they're so often focused on the speakers' personalities and windy, undirected things. I've tried to listen to so many podcasts and always dip an episode or two in because they devolve into people just chatting instead of actually presenting well-organized facts about what I want to listen to / learn about.

The structure and bare-minimum "human" aspect of this seems perfect for people like me to actually get into podcasts. I do wish I could further cut out all the disfluencies (um, like, uh, etc) though.

The only barrier for me IMO is wondering how accurate those facts actually are (typical research-with-AI concern).

I'm very much looking forward to a more interactive form of this, though, where I can selectively dive deeper (or delve ;) ) into specific topics during the podcast, which is admittedly very surface-level right now.

Re: NotebookLM's automatically generated podcasts are surprisingly effective

#405

Post by cjaackie »

mronetwo wrote: post #292
rafram wrote: post #283

Earlier quoted context omitted.

> The one and only point listening to a discussion about anything is that at least one of the speakers is someone who has an opinion that you may find interesting or refutable. No. Maybe that's true for you , but people enjoy learning in different ways, and some people learn best by listening to a discussion.

Unlikely. It's just that our brains are so fried by our smartphones/social media/24h of news/media consumption that we've lost the plot.

so that's the cool part, I think, instead of wasting time on socmed and news cycle composting, waste time on this instead. I think this is the general direction all media is headed, regardless of whether one agrees with it or not. Feed it whatever you want and it will shuffle together a plot, just for you.

Re: NotebookLM's automatically generated podcasts are surprisingly effective

#406

Post by quest88 »

chefandy wrote: post #394
quest88 wrote: post #385

Earlier quoted context omitted.

> Commercial creative workers are vulnerable because there's a billions-of-dollars industry effort to copy their professional output and compete with them selling cheap knock-offs. I agree there will be winners and losers of some proportion here. But I also think the people that want to pay for art will continue to pay as their motives and values are different. There's plenty of cheap knock-off art, but people still…

> But I also think the people that want to pay for art will continue to pay as their motives and values are different. The big difference is the type of artist. People selling fine art won't be affected much. The vast majority of artists are commercial artists, the the idea that being a commercial artist is morally or creatively bankrupt— a common sentiment among those who want to imagine that this is all just fine—…

I haven't seen that idea (artists being morally bankrupt), like you I'd strongly disagree. I also agree it's a shitty situation that artists invested hundreds of hours of their own time to create something only to be repackaged and sold by some AI tool.

That said, I'd still make the same point that people who value art and the artist will buy from and support the artist. Those that don't value it, won't. But now we're on a larger scale.

Re: NotebookLM's automatically generated podcasts are surprisingly effective

#407

Post by gcanyon »

AnIrishDuck wrote: post #401
gcanyon wrote: post #377

Earlier quoted context omitted.

> go is a game with an obvious score function which can be used to construct a loss, well defined moves, and total visibility of the board This is literally the opposite of true, and the main reason go computers were getting destroyed by humans for almost twenty years after deep blue took down Kasparov.

Yes, technically. But the broader point is true. Go is a game with well-defined win and loss conditions that can be automatically evaluated. This is critical for game-clock-eons of unsupervised self-play, which by most accounts is how AlphaGo (and other systems like AlphaZero) made the leap to superhuman levels of play. But it is entirely different from subjective endeavors like writing, music, and art. How do you sc…

Stipulating up front that this is a question for a lead scientist at OpenAI: I could see a scoring function looking at essays in the New York Times vs. the National Enquirer and finding a way to generalize from there. Similarly for the top 40 hit songs vs .

Re: NotebookLM's automatically generated podcasts are surprisingly effective

#408

Post by Terr_ »

llamaLord wrote: post #273
DanHulton wrote: post #254

Earlier quoted context omitted.

I keep seeing this asertion: "the robots will get there" (or its ilk), and it's starting to feel really weird to me. It's an article of faith -- we don't KNOW that they're going to get there. They're going to get better, almost certainly, but how much? How much gas is left in the tank for this technique? Honestly, I think the fact that every new "groundbreaking" news release about LLMs has come alongside a swath of d…

IMO it seems almost epistemologically impossible that LLM's following anything even resembling the current techniques will ever be able to comfortably out-perform humans at genuinely creative endeavours because they, almost by definition, cannot be "exceptional". If you think about how an LLM works, it's effectively going "given a certain input, what is the statistically average output that I should provide, given my…

IMO the problem facing us is not that computers will directly outperform people on the quality of what they produce, but that they will be used to generate an enormous quantity of inferior crap that is just good enough that filtering it out is impossible.

Not replacement, but ecosystem collapse.

Re: NotebookLM's automatically generated podcasts are surprisingly effective

#409

Post by mdp2021 »

InsideOutSanta wrote: post #305
mdp2021 wrote: post #295

Earlier quoted context omitted.

Since when industrial snacks are healthy food?

This probably isn't really a good analogy. It's just a fact that for most people, a conversation is more engaging than an academic paper. It's easier to pay attention to it, and it's easier to retain the information in it. This might be healthy food that tastes like a snack.

> a conversation is more engaging than an academic paper

I certainly agree with you, but it has to be quality conversation.

The example provided could suggest "think at what we could achieve" in an outcome that shows "and that is what could possibly go wrong".

Re: NotebookLM's automatically generated podcasts are surprisingly effective

#410

Post by mdp2021 »

1-6 wrote: post #373
mdp2021 wrote: post #295

Earlier quoted context omitted.

Since when industrial snacks are healthy food?

When I last checked, even healthy foodies occasionally enjoy shelf-stable snacks.

So let us say you could have effortful (as opposed to buttery-bread "no need to chew"), nutrient, and appetizing: if the snack is effortless (but for the bad spice) but with a hint of possibly no nutrients (when not possibly unhealthy), and tastes the-bad-way weird ("...like, OMG - I hmmmm was, like..."), where is the appetizing part?

If the "conversational form" (very good idea per se) has an implementation which would flow easily if not for the disturbing speech quirks, with doubts about the content quality: where can the interest be?

Post reply on HN