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

Why to Not Write a Book

gwern.net

Post reply on HN
Print view View original

191–200 of 215 posts

Re: Why to Not Write a Book

#191

Post by gwern »

gknoy wrote: post #181

Good grief, Gwern is a treasure. Every time I stumble across a link to something they've written, it's chock-full of side quest links to things that are _absolutely amazing_. I happened to leave my mouse briefly over one of the drop-caps, and it turns out there's an entire page dedicated to _generating drop-caps_ with AI, as well as showcasing all sorts of stuff. "Thank you for making such awesome stuff" seems so ped…

> I happened to leave my mouse briefly over one of the drop-caps, and it turns out there's an entire page dedicated to _generating drop-caps_ with AI

I'm very pleased to read that! I pushed Achmiz to implement that - "can we make all the dropcaps instances popup the https://gwern.net/dropcap page? "uh, probably, but why would you want to do that?" "because someone might be interested in one of the dropcaps and hover over it and otherwise they'll never know!" - and this is the first I've seen anyone mention it, but it sounds like it worked. It is one of the many touches few will ever notice, but I hope the ones who do will love it.

Re: Why to Not Write a Book

#192

Post by gwern »

taptak wrote: post #123
modernerd wrote: post #61

Earlier quoted context omitted.

> What happens when you walk up to a publishing agent and tell them you don’t have anything to show them… They told him to hand over whatever he'd got: "Douglas happily built his reputation for (at this stage, merely) sailing close to the wind when it came to publishing deadlines by telling interviewers that he had been ordered to literally stop writing and hand his manuscript over to a courier no matter where he had…

> "All of my life I've been attracted to the idea of being a writer, but like all writers I don't so much like writing as having written. … each time I meant to try to write something, I'd miss the deadline by two weeks." Having written several works without ever completing or publishing any of them, I find this comment from Adams particularly resonates with me. That said, I’ve had success with a narrative podcast, w…

Digital Antiquarian has a few posts up recently about Douglas Adams, and it all sounds like it was a great deal more painful than I had ever realized as a kid reading _Hitchhiker's_: https://www.filfre.net/2024/07/the-later-years-of-douglas-ad...

Considering how much he loved everything else and how well he was able to do things like radio shows or video or inspire text adventures, I do have to wonder if forcing him to write books was pounding a square peg into a round hole.

Re: Why to Not Write a Book

#193

Post by bonoboTP »

ghaff wrote: post #190
bonoboTP wrote: post #178

Earlier quoted context omitted.

Yes, I'm not trying to disparage it too much, it's just the reality of the job. But I found that even junior PhD students can have a romantic illusion that real scientists just spend their time musing and thinking and sometimes they have a flash of insight which they then enthusiastically share with the others. But the more mundane reality is that papers are projects with their own life cycle. You have to be on time,…

I've even had this discussion with people in at least adjacent roles to mine. We could put together ten very serviceable conference submittals tomorrow that we could spend a week turning into very competent conference presentations within a week. They wouldn't be brilliant but they'd be "good" relative to the norm. It's just one of the things we can crank out like we crank out blogs/columns.

Depends of course on the field and the conference tier. I doubt you could write an accepted paper for ICLR/ICML/NeurIPS/CVPR/ICCV in a week from scratch. These typically involve writing code, doing experiments, evaluating on benchmarks and obtaining good, if not state of the art, scores.

Typically this does involve 4-6 months of work even in high-output labs. It just involves many iterations, polishing the text and figures to "look like" an accepted paper by gestalt etc.

I wasn't trying to imply there's no work or effort needed. I'm more saying that there's a lot of learnable, repetitive parts and if you get them right, your output can have a consistent good quality but nothing groundbreaking. Chipping away at problems, permuting the assumptions and tasks, throwing in the latest hot new method and combining it with another method etc. It isn't something that anyone could do, but it doesn't rely on big eureka moments or flashes of deep insight that one might romantically expect.

Re: Why to Not Write a Book

#194

Post by ghaff »

bonoboTP wrote: post #193
ghaff wrote: post #190

Earlier quoted context omitted.

I've even had this discussion with people in at least adjacent roles to mine. We could put together ten very serviceable conference submittals tomorrow that we could spend a week turning into very competent conference presentations within a week. They wouldn't be brilliant but they'd be "good" relative to the norm. It's just one of the things we can crank out like we crank out blogs/columns.

Depends of course on the field and the conference tier. I doubt you could write an accepted paper for ICLR/ICML/NeurIPS/CVPR/ICCV in a week from scratch. These typically involve writing code, doing experiments, evaluating on benchmarks and obtaining good, if not state of the art, scores. Typically this does involve 4-6 months of work even in high-output labs. It just involves many iterations, polishing the text and f…

Certainly, if it needs to be some original/novel research that you've personally done of some sort, you can't just conjure that from scratch. A lot of what I've done over the years isn't that though.

Re: Why to Not Write a Book

#196

Post by wizzwizz4 »

Filligree wrote: post #102
Cthulhu_ wrote: post #83

Earlier quoted context omitted.

> He writes crime novels. Doesn't even like them, but if you want to sell, you either write smut (big gamble) or crime novels (less crowded). Reminds me of Chuck Tingle, who is an author but also a troll, writing books like "Pounded In The Butt By My Handsome Sentient Library Card Who Seems Otherworldly But In Reality Is Just A Natural Part Of The Priceless Resources Our Library System Provides".

Chuck Tingle is a treasure. On the surface all the books are smut, though this is actually because they're smut, but once you dig a little you'll find a ton of other amusements.

All of them are smut? I think you've forgotten Absolutely No Thoughts Of Pounding During My Fun Day With This Kind T-Rex Because I'm Aromantic And Asexual And That's A Wonderfully Valid Way Of Proving Love Is Real.

Re: Why to Not Write a Book

#197

Post by ghaff »

gwern wrote: post #192
taptak wrote: post #123

Earlier quoted context omitted.

> "All of my life I've been attracted to the idea of being a writer, but like all writers I don't so much like writing as having written. … each time I meant to try to write something, I'd miss the deadline by two weeks." Having written several works without ever completing or publishing any of them, I find this comment from Adams particularly resonates with me. That said, I’ve had success with a narrative podcast, w…

Digital Antiquarian has a few posts up recently about Douglas Adams, and it all sounds like it was a great deal more painful than I had ever realized as a kid reading _Hitchhiker's_: https://www.filfre.net/2024/07/the-later-years-of-douglas-ad... Considering how much he loved everything else and how well he was able to do things like radio shows or video or inspire text adventures, I do have to wonder if forcing him…

I know someone who presumably worked very closely with him on a text adventure game. I'll have to ask what it was like next time we talk.

Re: Why to Not Write a Book

#199

Post by gradschoolfail »

bryanrasmussen wrote: post #120
082349872349872 wrote: post #94

Earlier quoted context omitted.

Living in a high income country, where $200 is well within monthly fluctuation, this situation makes me wonder if a little more telepresence might lead to us bifurcating into a world where the Eloi will know a tiny amount about how all of it works, and the Morlocks will know all about how a tiny amount of it works?

isn't this how it already works?

Designers* will know all about how it is all supposed to work

*personally prefer the term “designists”, but YMMV on that one

Re: Why to Not Write a Book

#200

Post by a_bonobo »

dsr_ wrote: post #172
rusticpenn wrote: post #80

Earlier quoted context omitted.

Are you being sarcastic? The Robot novels are basically crime novels with robots...

It's not sarcasm: the conventional wisdom was that an SF novel could not also be a satisfying mystery/detective novel, because the readers could not guess that Aldebaranians can see in ultraviolet, or any other authorial invention. Asimov's insight was that it was up to the author to play rigorously fairly: every fact of consequence needed to be revealed naturally.

And that's what I love about the robot stories. He sets up the law of robotics, just like Agatha Christie and friends set up the detective fiction commandments, and then Asimov sets about finding all the loop holes in the laws of robotics. Every story is one loop-hole.
Post reply on HN