I heard of many conspiracy theories regarding COVID vaccines and 5G phone signals. People concerned that their very thought processes might be usurped by interlopers.
I haven't heard any such concerns about LLMs. Which are literally here to unburden you from your own thought processes. There's even a name for it: "Cognitive surrender." The LLM companies even embody the apocryphal drug-dealer stereotype: "First taste is free."
Retro photography hobbyists will explain why the most outstanding pictures seem to happen on film: Because the use of film affects the process of taking photographs, it slows things down and encourages the photographer to make a larger effort.
Recording engineers will explain why old recordings sound more vital and alive: Because the use of relatively-inflexible old analog tape systems affects the process of recording music, it encourages the in-person collaboration and focus of all involved.
Do you suppose that in the future, software developers will be explaining that old software catered to actual human needs and wants better? Because human authorship affected the software development process, that it required developers to consider their decisions more carefully? That it made fixing things more attractive than throwing things away and starting over wholesale?
@jonny And this is kind of bonkers. Because architecture has always been the hard part, the important part, of making big serious software systems.
Bad architecture is implicated in so many of the bugs that bite us, the security holes that betray us, and the technical debt that makes important projects ossify.
Architecture can be hard to see at a glance, impossible to appreciate without direct experience. One of the challenges of building complicated projects is just getting the stakeholders to take architecture seriously.
And now we have built this code-emitting machine that has no concept of architecture. Somehow, the stakeholders of the world are drooling. Because it promises to make irrelevant the things the stakeholders never understood.
/fin
@jonny There's been a theme to his discoveries, well exemplified by this post:
https://neuromatch.social/@jonny/116350918703314086
The LLM can write correct and coherent _lines_ of code. It can even write functions that appear to be okay in isolation.
What the LLM doesn't seem to be able to do is create any kind of coherent or comprehensible large-scale _architecture_ for a project as a whole.
2/
Mastodoner @jonny has decided that understanding the badnesses inside Claude Code is more valuable than his own sanity, and if you're interested in the writing of software and/or the future of writing software, you should follow him, so that the impending sacrifice of his aforementioned sanity will not have been in vain.
1/
Wife has job complaints. Her boss has a habit of telling her that her reports require more detail, excruciating detail. Then complaining that the revised reports are too verbose. Infuriating.
Her employer has recently signed up for a corporate-level Microsoft CoPilot subscription. Boss has recently encouraged her to engage CoPilot’s help, seeking brevity.
In this light, it sounds very much like boss has merely been running her reports past CoPilot all this time, and basing his feedback on its output.
She doesn't know what's going on for sure, she hasn't used CoPilot herself at all, and yet her trust has been undermined already, just because of the shadow of this thing.
I've seen some observers characterize the savage Vanity Fair photo-shoot of the assholes in the White House as "journalistic #photography" and that's just not accurate.
Those are #editorial photos, they illuminate a specific point of view, they tell the story the creators wanted to tell. They are mean, they are unfair, they are brilliant. They are in the photographer's signature style, but they do not extend any of the cover and assistance he might extend to someone he liked or wanted to please.
The fact that the administration people volunteered for this is a message of its own.
https://www.vanityfair.com/news/story/vanity-fair-goes-to-the-white-house-trump-2-edition
I tried getting ChatGPT's feedback on this William Eggleston photo, and the machine was (to my surprise) entirely negative. Its critiques were not crazy: "Composition Feels Loose," "Lack of a Strong Visual Hierarchy," "Scene Risks Looking Accidental Rather Than Intentional." (Aren't an awful lot of photos made to record scenes that are "accidental rather than intentional?)
But all of that could all be summarized as "This doesn't conform with the conventional wisdom of what 'good' photography is."
In this task, at least, I feel ChatGPT is akin to a dancing bear or a singing frog: Enormously surprising and impressive, but perhaps not as much use as the hype men were hoping for.
/fin
I thought I'd throw it for a loop with something really abstract, but it surprised me by identifying rust and a "cutout" in the surface, which indeed, it is. It did describe the un-rusted parts on the left as "muddy lighting," though.
3/
I tried it on a few of my pictures. I didn't think the feedback it offered was very useful, but I was kind of staggered by its ability to recognize things in a photo. It identified this as including weeds, machinery, metal, and a tire. It didn't identify a tractor, though.
2/
Internet-famous photographer Sean Tucker tries asking ChatGPT for feedback on his photography.
I hadn't realized this was possible. The idea makes me uncomfortable. Now LLMs aren't just threatening entry-level employment opportunities, they're branching out into threatening traditional places for this sort of artistic and aesthetic feedback: clubs, meet-ups and other kinds of in-person and online social activities.
Even if we give the machine every benefit of the doubt, won't it, at best, be comparing our work with a kind of societal average? Won't it just be telling us what would make our photos more popular on Instagram?
1/
Huh, actually, Cloudflare's "Global Network" page lists my city, Cleveland, and several close neighbors, like Columbus, Detroit, and Indianapolis, as data center locations.
Maybe the issue is slightly too complicated for an amateur to suss out in five minutes. I mean, possibly. Stranger things have happened.
I opened up the web for my daily coffee time and found MetaFilter, Kottke.org, The Old Reader, and even media.universeodon.com are all inaccessible to me, because of some sort of Cloudflare error in (depending on the error message) Chicago or Toronto.
I am hundreds of miles away from both of those cities. Surprised there isn't some sort of Cloudflare infrastructure closer to home.
In the early days of the internet, optimists and futurists lauded the internet for its fault tolerance. Things have changed.
Little websites (like my Mastodon instance) that haven't tied themselves to Big Internet services are working, though.
"A List Of Text-Only & Minimalist News Sites" is like a peek into an alternate timeline, where the web evolved in a different direction. Where it became functional and friction-free instead of bloated and annoying.
@dirtyoldtown got me wondering about buy-now-pay-later service Klarna, and I found more than I was expecting.
https://mefi.social/@dirtyoldtown/115442625013754652
Back in 2015, Swedish Minister of Enterprise Mikael Damberg referred to five Swedish companies as the "five unicorns." He meant it in a flattering way. Those companies were Spotify, Klarna, Mojang, Skype and King.com.
Skype is gone; Spotify has platformed Joe Rogan and separately been criticized as a platform for money-laundering; Klarna is in the business of getting Gen-Z into debt over bagatelles and profiting from their disorganization, and King.com takes in billions with casual-gaming micro-transactions.
That's ... a surprisingly seedy lineup. Coming from Sweden. I'm surprised, anyway.
It's like reading a Carnegie-era paean to the industrial might of Pittsburgh, only to find out later that it was really mostly pawnbrokers.
I went to #NoKings. In a suburb. In a place all my family could get to. It was many times larger than I ever would have imagined.
"If Melania doesn't have to live with him, why do we?" was a quip I hadn't seen before.
"Eggs are expensive because the chickens are all in Congress" was another standout.
Science and electronics YouTuber Fran Blanche alleges that YouTube has turned the volume down on trans advocacy by refusing to recommend any sort of positive videos on the subject.
If I watch woodworking videos, YouTube suggests more. If I watch videos about video games, YouTube suggests more videos about the same game. If I watch videos about tools or cars, YouTube suggests more of those, plus firearms. YouTube is definitely paying attention to the subject matter I watch.
I've just watched a bunch of videos about trans issues, analysis of trans critics, and trans rights, and YouTube is conspicuously not offering to show me more. As if it didn't notice.
I think Fran's conclusion checks out. YouTube can give you a good introduction to a million different topics, and it's great for how-to-do-this demos. But the recommendation system has long held potential as a conduit for invisible, insidious influence.
There's a common format for human interest news stories now, and they could all have the same headline:
"Can this person, who voted for TFG and got screwed, at least admit that supporting him was a mistake?"
The answer is always "Ha ha no of course not."
What is the point, what is the purpose of these stories? What is illuminated, who learns anything from this?
A fellow could almost get the idea that the point is to squash hope, and to demotivate any sort of action.
Sometimes we hear hand-wringing over young people, not having relationships, not getting married, not having kids. Could this not be explained, to a large extent, explained by the deliberately-engineered environment of confusion and hopelessness those young people have had to contend with for half of their lives? The "Russian Firehose of Falsehood" is an emotional attack, it could have emotional effects, go figure.
The hand-wringers blaming "social media" and "cell phones" for these trends never seem to talk about the grim and depressing emotional landscape we're ALL faced with every day, even those of us who don't even own a Toktick or whatever.
