DynamoSunshirt's recent activity
-
Comment on The rise of Whatever in ~tech
-
Comment on The rise of Whatever in ~tech
DynamoSunshirt Link ParentThanks for this. In my case, I was just curious about the comparison to another body of water, so I wanted a quick estimate, not a piece of code. Even 15 minutes is more than I was hoping for. But...Thanks for this. In my case, I was just curious about the comparison to another body of water, so I wanted a quick estimate, not a piece of code. Even 15 minutes is more than I was hoping for.
But I do appreciate you breaking down your thought process here. It seems like you are indeed using LLMs the only way that I've found them to be useful: as a starting point. Except you're definitely better about iterating on a prompt than I am. I feel like I hit walls much more often when I try to call out mistakes and inaccuracies -- the LLM will politely apologize for the misconception and claim to fix it, but the vast majority of the time it gets stuck in a cycle of introducing new bugs while only partially resolving the past mistakes. I've been burned too many times by that while under the gun to complete a task at work, which ironically makes me less willing to experiment with LLMs since I know I could likely solve the problem myself in less time if I don't faff with the LLM in a neverending spiral of lies.
One takeaway that perhaps others can benefit from: it seems that suggesting a possible solution, like you did with OSM and a Python script, is always the way to go with an LLM. That makes it tough for me to navigate spaces that I'm not already familiar with, because I tend not to try to 'solution' when I ask people questions -- I tend to want to defer to their expertise! But with an LLM, I should remember that it's a model, not a person, so I'm not wasting its time by steering it. Same goes for iteration; even if I don't fully understand the codebase, better to speculate about the possible bug and a solution for it with an LLM. Whereas doing that from an uninformed perspective with a human being who slapped some code together would typically be pretty annoying.
-
Comment on The rise of Whatever in ~tech
DynamoSunshirt Link ParentThat's how they all work in my experience. Please provide some sources if you're going to claim that other models have advanced past this fundamental limitation of LLMs. I found a recent Apple...That's how they all work in my experience. Please provide some sources if you're going to claim that other models have advanced past this fundamental limitation of LLMs. I found a recent Apple research paper on this exact subject to be a great read.
-
Comment on The rise of Whatever in ~tech
DynamoSunshirt Link ParentIf present-day LLMs showed up in a classic Trek episode, I can only imagine two use cases: an episode where the ship computer becomes infected with malware that turns it into a pathological liar...If present-day LLMs showed up in a classic Trek episode, I can only imagine two use cases:
-
an episode where the ship computer becomes infected with malware that turns it into a pathological liar in some sort of dastardly plot to destroy the ship
-
an artifact of a failed, post-apocalyptic civilization destroyed by its own hubris
Come to think of it, I would absolutely watch an episode about those two concepts combined...
-
-
Comment on Amazon now counts more than one million robots at its facilities in ~tech
DynamoSunshirt Link ParentI feel the same way about ride-sharing apps (or, as we used to call them before the Business Idiots rebranded... taxis). The work seems exhausting. Sitting down in a car for many hours a day,...I feel the same way about ride-sharing apps (or, as we used to call them before the Business Idiots rebranded... taxis). The work seems exhausting. Sitting down in a car for many hours a day, unable to get up and stretch at all, is awful for you. Just finding a place to go to the bathroom seems like a big obstacle. And with the 'gig economy', a lot of people severely underestimate their personal costs in terms of wear and tear and unpaid time between rides, leading people to think they're making a lot more money than they really are. We are probably better off if all of those jobs get automated away.
But as other commenters point out: the more entry-level jobs you automate, the fewer jobs there are for folks who aren't interested in a career (for whatever reason). We did the same thing to housing when we got rid of halfway houses, the same thing to food service when we lost cafeterias and automats, and the same thing to transportation when we made it impossible to walk to most places in the USA. You can eliminate the bottom of a lot of industries, but if you do that, you need to simultaneously lift up everyone so nobody is left behind. If you kill the halfway house, everyone needs to get on leases for apartments as a bare minimum. If you kill walkability, everyone needs to buy a car.
You only have to visit the entire West Coast of the United States to see what happens when you leave people behind. They don't just disappear.
-
Comment on The rise of Whatever in ~tech
DynamoSunshirt Link ParentWell said. I would find LLMs maybe possibly almost useful if they could reliably say "I don't know." Funnily enough, they share this weakness with most of the worst hype bros I've worked with in tech.Well said. I would find LLMs maybe possibly almost useful if they could reliably say "I don't know." Funnily enough, they share this weakness with most of the worst hype bros I've worked with in tech.
-
Comment on The rise of Whatever in ~tech
DynamoSunshirt LinkI wanted to raise the visibility of these paragraphs because this is my exact experience. As soon as you try something that nobody on the internet has tried before, the LLM makes shit up....- Exemplary
The trouble was likely that there was no built-in way to do what I wanted, and no one had ever successfully done it before, so the machine had nothing to draw from… and simply generated something that sounded plausible instead. Because that is what this technology does: it continues a conversation in a way that sounds plausible, as defined by similarity to existing conversations. If there are existing conversations about the topic, great! That makes for a more specific measure of plausibility. If not, even better! Just about anything might be plausible! It can just generate Whatever!
I cannot stress enough that this is worse than useless to me. Not only did it not answer my question, but it sent me on a wild goose chase making sure I had not somehow overlooked the fake API it generated.
I wanted to raise the visibility of these paragraphs because this is my exact experience. As soon as you try something that nobody on the internet has tried before, the LLM makes shit up. Sometimes, with programming boilerplate boring API plumbing, this is highly predictable and the LLM nails it.
Most of the time in my experience, it doesn't have any useful input. For instance, I tried to get Claude to calculate the surface volume of a body of water near where I grew up. Wikipedia, nor anywhere else, has that calculation. So it literally just grabbed an acreage number from a real estate listing of a property nearby.
When LLMs spew these fake facts -- and they often do -- I feel like I'm polluting my mind. I don't want to spend my time fact checking an LLM, even if it gives me citations, because I can just web search for that content myself and skip the slop.
It makes programming spaces feel bleaker. I don’t want to help someone who opens with “I don’t know how to do this so I asked ChatGPT and it gave me these 200 lines but it doesn’t work”. I don’t want to know how much code wasn’t actually written by anyone. I don’t want to hear how many of my colleagues think Whatever is equivalent to their own output. I don’t want to keep watching people fall for a carnival trick.
We did a hackathon at work recently. Every project? An AI chatbot, bolted onto a different facet of our product. Whatever, lame, but I thought it still might be cool to dive into unfamiliar codebases and learn some of our frontend code.
Except everyone was just vibecoding with Cursor because they didn't know the codebase. This method let them sling thousands of (garbage) lines of Whatever code that kinda sorta did the job and only broke 10% of the time. And despite not understanding the codebase, the chatbot API (just try asking any of them how to maintain context), or the code they generated, every demo was incredibly braggy and acted like they crafted this output themselves and understood it all. Most haven't even read more than a couple of their own lines.
An intern asked me 'What were hackathons like before LLMs' and I nearly cried.
LLMs (which are most certainly not sci-fi AI, with any level of independent thought or conscience, hence the constant boldfaced lying) have sucked all of the air out of the room on so many subjects. Not just programming, but now teaching, most online writing, journalism, even reviews for products. I have given them a chance, deluding myself into thinking I should keep an open mind since they have so much potential.
But I'm glad I read this article because it has convinced me to finally stop giving LLMs a chance. Fool me once, shame on me. Fool me twice -- nope, I am just done listening to these garbage-spewers. I only hope the hype wave crests soon enough that I'll have literally anything else to talk about online in a couple of years.
-
Comment on What is your opinion whenever you see news/opinion that tech companies are relying more on chatbots rather than junior developers/interns? in ~tech
DynamoSunshirt Link ParentGreat point. I tend to use Kotlin, Python, and Swift a lot, and my frontend work is very basic compared to React.Great point. I tend to use Kotlin, Python, and Swift a lot, and my frontend work is very basic compared to React.
-
Comment on What is your opinion whenever you see news/opinion that tech companies are relying more on chatbots rather than junior developers/interns? in ~tech
DynamoSunshirt Link ParentThank you for the recommendation, but I have already tried agent mode with a number of different models. For my work, it simply doesn't meet my needs. I generate most boilerplate automatically....Thank you for the recommendation, but I have already tried agent mode with a number of different models. For my work, it simply doesn't meet my needs. I generate most boilerplate automatically. The rest of my code tends to require attention and deep thinking.
I think a lot of cookie cutter REST API jobs can benefit from LLMs just to ease the monotony, especially test writing. But when you write something comprised mostly of algorithms and your code needs to be highly optimized, I find LLMs come up short.
FYI, this 'but you need to try AGENTIC mode in <trendy new IDE>' mentality is exactly the kind of spammy nonsense I'm talking about. I have tried Cursor. I have tried VSCode with copilot. Extensively in the last few weeks. Why can't LLM enthusiasts understand that it just isn't that useful to me in my workflows? I'm glad it works for you. But LLM enthusiasts really need to understand that not everyone needs or wants to use the same workflow as you. Diversity is good.
-
Comment on What is your opinion whenever you see news/opinion that tech companies are relying more on chatbots rather than junior developers/interns? in ~tech
DynamoSunshirt Link ParentI can only disagree. Sure, LLMs can be useful, especially in experienced hands. But I tried to keep an open mind and use them and I just do not find them particularly helpful. The problem isn't...I can only disagree. Sure, LLMs can be useful, especially in experienced hands. But I tried to keep an open mind and use them and I just do not find them particularly helpful.
The problem isn't the tool: anyone who wants to use LLMs should be able to. The problem is management at many companies, who increasingly believe the lies that LLMs are equivalent to a junior developer and that all of your devs need to embrace LLMs to 10x their output. Mandating and forcing LLM usage is making my job (and I suspect many others) totally miserable.
If I was an author, and my publisher tried to force me to use LLMs to make writing books go 5x faster, I would get a new publisher. I feel exactly the same way about my current employer.
-
Comment on What is your opinion whenever you see news/opinion that tech companies are relying more on chatbots rather than junior developers/interns? in ~tech
DynamoSunshirt LinkShortsighted greed led by gullible managers who have completely lost touch with what actual work is like. I find the modern tech 'AI' obsession completely soul-crushing. It can be useful for...Shortsighted greed led by gullible managers who have completely lost touch with what actual work is like.
I find the modern tech 'AI' obsession completely soul-crushing. It can be useful for generating ideas and rubber ducking. Most vibe coders are self-promoting lying charlatans or are just bad enough at coding that they think 'AI' is producing a decent result.
We've lost sight of what makes work worthwhile. We've lost sight of what makes a product or company worthwhile.
Hopefully the 'AI' (no, it is not intelligent, nor is it the AI I know from sci-fi stories, hence the quotes) craze starts to collapse under its own weight soon. At this point I hope it takes the whole corrupt, bloated, immoral tech industry down with it. Which I suspect it might either way, if these companies keep threatening to fire people who don't vibe code their way to 'efficiency'.
-
Comment on Laser-wielding device is like an anti-aircraft system for mosquitoes in ~tech
DynamoSunshirt Link ParentA fan also works quite well! Ceiling fans are best in my experience. I imagine that, to a mosquito, a ceiling fan on high is basically a tornado.A fan also works quite well! Ceiling fans are best in my experience. I imagine that, to a mosquito, a ceiling fan on high is basically a tornado.
-
Comment on AI videos have never been better: can you tell what's real? in ~tech
DynamoSunshirt Link ParentVery convincing. But I do have to question how hard it is to fake compressed short-form low-quality videos. I still haven't seen any AI-generated videos that seem convincingly like a movie. And of...Very convincing. But I do have to question how hard it is to fake compressed short-form low-quality videos. I still haven't seen any AI-generated videos that seem convincingly like a movie. And of course the short-form video is a great way to work around the limitations of limited context and persistence; right now, if a video spins 360 degrees and the scene stays consistent throughout, you know 99% confidence that a video is real. The same thing applies to longer form content in general; it's easier to spot inconsistencies if a scene goes on long enough that inconsistencies can actually happen. 10 seconds is simply not enough time.
This is really really really scary considering how many people consume tons of short-form video. Many people do not consume content through a critical lens. Many people get news from short-form videos on TikTok and Instagram. I've already noticed that these people essentially live in a different universe from myself, believing whatever the hottest algo influencers are pushing in a given week. Surely this will make that worse, and easier to manipulate than ever -- no need for companies, governments, etc. to pay off an influencer to create propaganda for you when you can just generate it yourself through a hot influencer-ish AI avatar.
I continue to feel that short-form video is the lowest form of communication content available today, right next to image memes. Not enough length to give citations or make a coherent argument with any nuance. Reliant on eye-grabbing techniques just like YouTube previews. I call these 'eyeworms' because they remind me of music earworms, and yeah, I want the term to sound disgusting. But honestly this stuff is getting creepily similar to TV advertisements -- no substance, just trying to sell you something or brainwash you in 10-30 seconds. And unlike TV ads, there's no regulation whatsoever. So you don't even get a giant screen of tiny text sideeffects read in a super fast voice when someone tries to sell you drugs on TikTok.
Back to RSS feeds and blogs and obsolescence I go.
-
Comment on Contra Ptacek's terrible article on AI in ~tech
DynamoSunshirt LinkI read Ptacek's article a few days ago, and it really bothered me. This article is a (well-written) rebuttal that explores some of the fallacies in the original article. If you're feeling bad...I read Ptacek's article a few days ago, and it really bothered me. This article is a (well-written) rebuttal that explores some of the fallacies in the original article. If you're feeling bad about 'not taking full advantage of genAI' these days (even if you deep down don't want or need to, the societal pressure is insane in some fields right now), this might make you feel better. And give you a chuckle.
-
Contra Ptacek's terrible article on AI
27 votes -
Comment on What's the most feasible way to exit modern society? in ~talk
DynamoSunshirt LinkI tried to escape. Moved to a small town. Lived locally. Picked up a lot of outdoor hobbies. Made a whole load of new, local friends. But reality catches up anyway. I suspect you could outpace it...I tried to escape. Moved to a small town. Lived locally. Picked up a lot of outdoor hobbies. Made a whole load of new, local friends.
But reality catches up anyway. I suspect you could outpace it for longer if you built an off-grid self-sufficient cabin or something... but you'd need to either be 100% introverted, or convince friends to come along. After a few months, it's easy to want to return to society if you start getting bored. And ultimately you can't cut yourself off completely; you'll always need to go to the doctor, dentist, barber, grocery store, and plenty more.
I also loathe genAI. I tried to give it a chance and have an open mind, but ultimately I just find it a lazy shortcut that generates garbage. I'd rather sit and think about my problems. But I also think we're close to the peak of a fad, so I have hope that we might limit genAI to more appropriate use cases in the next few years as the VC capital fades and the cost rises and enshittification ensues. What makes you loathe it so much more?
-
Comment on Hogwarts Legacy and designing games for the masses in ~games
DynamoSunshirt Link ParentAnd they're all just ripoffs of the Assassin's Creed quest-and-task-stuffing time waster concept.And they're all just ripoffs of the Assassin's Creed quest-and-task-stuffing time waster concept.
-
Comment on Weekly US politics news and updates thread - week of June 16 in ~society
DynamoSunshirt Link ParentI'm hoping this gets him the attention he needs to beat famed sex pest, narcissist, and murderer of the elderly Cuomo. But it is hard to predict how people really feel about immigration these...I'm hoping this gets him the attention he needs to beat famed sex pest, narcissist, and murderer of the elderly Cuomo. But it is hard to predict how people really feel about immigration these days. Even educated, typically thoughtful people I know quite well seem to have been radicalized into thinking that all immigrants are e-bike riding societal leeches who won't stop endangering pedestrians.
I think a lot of people in the city are upset about migrant hotels as well. And perhaps justifiably; there are a lot of New Yorkers who struggle to get by, often winding up homeless. It's easy to see how they might take free housing and supplies for illegal immigrants as an affront, even if I think helping both groups isn't mutually exclusive.
-
Comment on Coco Robotics raises $80M to scale up autonomous delivery fleet in ~tech
DynamoSunshirt Link ParentEven worse, they'd disrupt bike traffic. Bike lanes exist to provide a safe environment for unshielded humans on dangerous streets. Autonomous delivery vehicles can't die. Toss them into the street.Even worse, they'd disrupt bike traffic.
Bike lanes exist to provide a safe environment for unshielded humans on dangerous streets.
Autonomous delivery vehicles can't die. Toss them into the street.
-
Comment on A huge outbreak has made Ontario the measles centre of the western hemisphere in ~health
DynamoSunshirt Link ParentHonestly you could cover 99.9% of it with a requirement for school sports. And better yet, the kind of people who oppose vaccines have a really strong overlap with people willing to do whatever...Honestly you could cover 99.9% of it with a requirement for school sports. And better yet, the kind of people who oppose vaccines have a really strong overlap with people willing to do whatever Coach says to Win the Big Game. Even if it means poking a needle in their precious, unspoiled, not-yet-dead-from-measles child.
All true, but what I was mostly referring to was the vibe. It's a lot weirder to hang out with people around a table where everyone feeds input into Cursor than the vibe 10 years ago, where everyone would frequently gather around one or two laptops (often one for frontend, one for backend) and collab.