@cadey Looking at Google’s SEO changes recently (ignoring their AI search stuff which seems to be driven by MS getting to it first), I get a feeling that they’re aware of the impossibility of this problem and their solution to moderating the LLM sludge is to… push the moderation to others.

More specifically, it seems they’re starting to push social media and forums more. They added “discussions and forums” card to some searches, added a “perspectives” filter that’s supposed to, again, highlight “organic” discussions, and yesterday finally added support for forum-related structured data (ProfilePage and DiscussionForumPosting, though the latter seems to have some parsing issue with microdata [unless I misunderstood how microdata should handle <a> elements]).

Their updates to Search Quality Raters Guidelines also seem to indicate that they’re trying to move search results more towards “user experiences” as more and more articles are AI-generated…

The obvious issue is that social media and forums are also spam targets, and obviously malicious actors can try just setting up entirely fake forums too, but on the other hand moderating a smaller silo is much easier, and while it’d also have some negative effects, enshrining existing forums/social media somewhat by promoting them over newer ones would at least increase difficulty of such operation.

Still, these moves feel very much like passing the ball to social media and forum moderators instead of actually tackling the issue.

Yes. This is exactly it. And, to quote Kerninghan:

Debugging is twice as hard as writing the code in the first place. Therefore, if you write the code as cleverly as possible, you are, by definition, not smart enough to debug it.

No, I know, Kerninghan didn’t live in a world with generative “AI”, but it’s worth considering the meaning of his words in this context…

If you think you can use generative models to write code, and especially if you think that you can dispense with coders because of generative models, then you’re going to have a really bad time when it comes to debugging. You’ll have to understand the complete system, as Jennifer points out here, you’ll have to understand all the code that the “AI” wrote for you so deeply that you understand the ways it could fail, and you’ll have to have talent in a suite of tools that make debugging certain aspects possible.

I know that a lot of people are saying that generative AI will “just keep getting better” and “will stop writing buggy code”, but I don’t think that’s possible for at least two reasons.

1. It’s trained on a lot of truly shitty code. (imagine every half-baked first-year student project left on github, and bad stack overflow answers)
2. Again, as Jennifer points out, it assumes that the “prompt engineer 🥴” knows the whole architecture well enough to ask for the right code in the first place.

Another successful downs-only #nytxw solve!

In hindsight though, I now wish I had kept track of them. I didn’t really think I’d reach the point where I’d do enough to lose track. (Probably 5ish now?)

@zenspider Neat, though I don’t think I’m committed enough to this idea to actually try it. 😬

@zenspider Probably, it’s not something I’ve actively looked into a solution for since I’m on macOS anyway.

Going to be on the east coast for the start of #AdventOfCode this year, so it’s going to be extremely unlikely I’ll be awake to attempt leaderboarding. 😢

And since I’ll be away, it also seems unlikely that I’ll have the time to do solutions in anything but Ruby, but I’m hoping to give Uiua a shot for some of the early problems this year.

What I want a surprising amount of the time is a FUSE view into a database so I can have more structured data but still easily browse/edit with tooling of my choice without having to be constrained by actual filesystem representation of directories and files.

@philcrissman Reading through @zenspider@ruby.social’s zenweb source was pretty interesting for seeing how he approached some of those complexities.

Watching a mutual ask for printer recs and receive a chorus of tired tech folk going "Just get a Brother, they're fine" and man

MAN

Like this is actually kinda fascinating honestly, Brother is now the best printer brand, the one that every Computer Person recommends, and is it because their printers are good? Their printers are fine, they print, whatever, no, it's because everybody else's printers have gotten Innovated out the wazoo, every innovation making them way worse, until it's gotten to the point where I wouldn't have one in the house even if it were free, and meanwhile Brother's have remained consistently Fine I Guess, which now makes them the best printer manufacturer simply by virtue of them opting out of the Who Can Get Crappiest Fastest race

Brother have gotten to where they are now, by NOT innovating

EDIT 2023-11-27 2130 UTC: I muted this thread a while back because wew lad this got too big. I won't see your reply but the lurkers (and commentors) of Hacker News might.

K, I apparently am late to the "sharks have been around for a ridiculously long time" thing. I mean, I know they're ancient, but I really did not have a clue just how ancient, relative to other "things that have been around forever".

The one that blew my mind is the age of sharks relative to the existence of the universe.

By our best understanding, the Big Bang happened about 13.8 billion years ago. Sharks have existed on Earth for about 450 million years. That means sharks have existed for approximately 3% of the entirety of the existence of the Universe.

Standard "some things that sharks are older than" list:

• Saturn's rings (~100 million years old)
• The Pleiades cluster (75-150 myo)
• One galactic orbit of the sun (200-225 my)
• North Star, aka Polaris (estimated at ~70 myo)
• The Atlantic Ocean (~150 myo)
• Pangea (formed ~335 million years ago)
• Trees (390 myo)

There are too many good anecdotes to quote in these notes on a UX vs. security in healthcare paper. You should just click through and read the whole thing.

https://ferd.ca/notes/paper-you-want-my-password-or-a-dead-patient.html

The "effective altruism" and "effective accelerationism" ideologies that have been cropping up in AI debates are just a thin veneer over the typical blend of Silicon Valley techno-utopianism, inflated egos, and greed. Let's try something else.

https://newsletter.mollywhite.net/p/effective-obfuscation

#ArtificialIntelligence #AI #AIEthics #EffectiveAltruism #EffectiveAccelerationism #newsletter #CitationNeeded

"Elon has lost his wife, his kids, 40 billion dollars, and his space ship crashed. It’s like a genre of country music that doesn’t even exist yet"

Once Tesla completes its Cybertruck and Full Self Driving technology, then Elon's truck will be able to leave him, too.

#ElonMusk #CountryMusic #SadCowboySong

So you’re telling me that arguably the most effective collective workers action in Silicon Valley history was in support of the billionaire CEO at OpenAI, a company founded to put people out of work

So many profile pieces in the media about MAGAs after Trump won in 2016. Did they become milder? No, they disgusting fascism just became mainstream. We know where this ends.

_They_ are the bullies, and _we_ get to decide if we want them at our birthday parties. And we don't. I have no empathy for those who lack it completely. 9/9

What the heck, a git subcommand that I didn’t know about?! https://git-scm.com/docs/git-notes

I thought I was long past the point of at least being aware of most normal stuff about git.

on the upside, found this very funny meme again

The NYT on a Georgism (or Geoism) proposal in Detroit: https://www.nytimes.com/2023/11/12/business/georgism-land-tax-housing.html?unlocked_article_code=1.90w.2d9n.Jzc050i-Hhei&smid=url-share

I came to this via thinking that it was absurd that people can profit off buying and selling land, and someone pointing me at Georgism, which was that, but so much more.

@josephholsten UniFi Protect?