We're entering the era of human slop, where anything made by humans is worse than AI.
>twitter screencap threadSNCA
AI doesn't exist.
>>107822787Giving access to the internet in 3rd world countries was big mistake
>>107822787>humans is worse than AIShitty bait threadAll fields
>>107822793fpbp
>>107822807But it's true
>>107822807Better than 99.99% of humans for sure. Claude is a better coder than every single junior working under me. I don't even delegate tasks to them anymore. It's literally more efficient to do all the tasks myself with Claude and faster as well.It's clear software engineers will face contraction soon because of this. I don't even know how other white collar jobs can keep existing when they are already this good. No way most people still do manual work instead of delegating it to AI right now.
>>107822787glowies kill yourselves, there is no such thing as AI
>>107822941It's everywhere now, look around you
>>107822928Is your code actually maintainable when it's made by AI? Are you sure you even understand how it works well enough to document and explain it to someone who would need to adapt it?
Humans will look pathetic very soon
>>107822928Idk what you are doing, but if you are doing some simple webdev shit then I am not surprised. You webdevs were automated almost a decade ago with bullshit like Squarespace and other template builders.
Kill every single Indian
We are about 90% done with automating a programmerThe next step is to automate testing and Q/A
>>107822797>>107822941Real Men know: "AI" is just IA.
>>107822787Humans have always produced slop. At least 90% of humans are dumb as bricks. AI produces slop because it was trained on the work of average humans.
>DeepSeek V4>GLM-5>MiniMax-M3>Qwen 3.5>Nemotron Ultra>Kimi K3
>>107822972He is a senior dev with devs working under him so I'm guessing he knows enough about he code to understand it lolAll of those issues you stated is the same if he were to delegate it to a junior dev. Except they're worse because you cant micromanage and hold a junior dev hand like you can Claude.
shouldn’t you be on the frontline?
>>107823107Is there russian ai?
>>107823118kandinsky, gigachat, yandexgpt
>>107823133That's wrappers
>>107823139not wrappers. cope & dilate
I'm ready
>>107823022Shalom
>>107822928so what are you working on?
>>107822972NTA, but yes, my code generated by ai is maintainable. First, I can understand the code myself so if AI were to stop existing, it would still be a net positive. Second, AI will only get better and more prevasive, so the maintenance will be even easier. Third, AI of the future will have code that will be unmaintainble by humans, because they will be infinitely more smarter than humans at coding. When that happens, my worries wont be coding, it will be feature outsets. "I want things to work, I dont care how.". It will be similar to how I view the image gen diffusion today myself. I'm not an artist, I will NEVER be an artist and get better at drawing an art of any magnitude like image gens can. So I dont care what techniques are behind an image output, other than that it just looks good, and better than what I can imagine myself seeing. "Its better than my imagination". That will be coding.
>We're entering the era of human slop, where anything made by humans is worse than AI.
>>107824011True
OP is proof.
>>107824011>>107824150trueth nuclear bomb
>>107823687
>>107824394
>>107822787What's going to happen when most of the white collar jobs gets replaced by AI? You can't expect all of them to be welders. Do we just keel over and die?
>>107824692
>>107824749
>blue checkmark>anime avatar>AI centered "social media influencer"
>>107822928My company is literally doing the opposite. They are hiring cheap juniors and asks them to implement core features. Switching my $DAYJOB is long overdue.>>107823990>infinitely more smarter than humans at codingI can't tell if it's bait or if you are actually retarded.
>>107824934
>>107823092> Except they're worse because you cant micromanage and hold a junior dev hand like you can Claude.My nigga, you can teach a junior dev how to not be useless. You can't teach Claude anything. The best you can hope for is that with enough "prompt engineering" it will accidentally stumble into something useful.Also, someone being a "senior dev" at some rando company doesn't make them competent. Remember, it was a senior dev that made the decision which took down cloudflare due to their rust shortcuts. >>107823990> Third, AI of the future will have code that will be unmaintainble by humans, because they will be infinitely more smarter than humans at coding. Brother, where can I get this hopium you're smoking? The most dangerous part about LLM's is that we have a whole industry of people who have confused mimicry of a world model for actually having a world model. These things literally don't have logical structure to their "thought processes" and the only way we can fake it is by making the autoregressive sequence of tokens for inference exponentially longer (which is entirely unsustainable). If we get AI that are "so smart they make unmaintainable code because humans are too dumb to understand it," it will be by humans becoming retarded and experiencing cognitive decline.
>>107825562>you cant teach anythingAll the AI models are better than they were 1 year ago, better than they were 2-3-4 years ago. Hell they didnt even exist prior to 4-5 years ago. So it came out of nothing from shitty chatbot to coding capability to smart agent like behavior that can create files, write codes, fix codes, compile, check errors, fix errors, etc. Behind the scenes in the AI labs, they have automated pipeline up the wazoo. You have no idea what you're talking about
>>107825593You are not understanding. The Claude that you interact with is literally not the same model or architecture as previous deployments.You can make model by model improvements, and in principle you can make marginal improvements to specific tasks via HFRL, but you can't actually "make Claude better at solving the problem" simply by interacting with them. You can do that with a junior dev. Simply by doing the work, they accumulate skills and improve at their job. That doesn't happen with a transformer model. No matter how many tokens you inference, if Claude is having trouble solving a problem, that can only be fixed by either retraining (risky due to the "forgetting" issue) or re-designing the architecture.
>Make a 10 word twatter post>Can't even be arsed to write it himselfThe absolute state, these niggers literally offloaded all their brain functions into a fucking AI...
>>107825640Continuous learning is a thing thats becoming a thing
>>107825593https://archive.4plebs.org/tg/thread/91378195/I would say the art here was about the same par of quality as the shit we have these days.
>>107825807SD has gotten really good at specialized training, and its my goto for generating art and porn. And these training datas have gotten lot better over the years too. However newer proprietary models have come out that are technically better, but without training for commoners, all you get is generalized model. But really, once you hit a upper limit of our own understanding, you cant tell the difference due to our own limitations.
>>107822787>according to two people with direct knowledge of the planThey sound impartial!Also,>Twitter screenshot>From a random nobody>пepeвecти пocтWhy would anyone care about this?
>>107825795This year, yes
>>107823990By the time AI becomes so good at coding we will not need human reviewer we will have AGI with 90% of jobs automated. Your AI generated code project will have no users either because there are no humans with enough disposable income to own a computer left, or because everyone generates things on the fly.>Better then my imaginationSomething like 50% of people can’t imagine apple with colors and everything, so raw quality is not a competition, and when it comes to something you want then there is literally nothing better then your imagination, because it’s your’s.
>>107822972NTA but yes? You can bake it into your prompt that you want extensive human readable, coherent documentation for all code it writes. This is something I do if I'm having a LLM write code but I'm not going to be able to directly supervise and watch what it's doing at every moment. Then I look over the code and documentation to determine whether or not I'm telling it "great job, here's your next task" or "fuckin' dumb clanker, rewrite this garbage, you fucked it."
>>107822972>maintainablemade-up meme metric
I hope AI will help reimplement game engines.
>>107822787Human slop has always existed.But slop is in the QUANTITY of shitty content, not just shitty content existing, and AI will always beat humans in that because you can produce it in seconds at the push of a button.
>>107822928It's going to be great when you leave your job.
>>107822787>anything made by humans is worse than AI.>we made the AI
>>107822787I've discovered AI isn't really good at anything except generating FOMO slop and drowning the internet in it.
>>107822928saasrs i thinkeing the same thing... cxlaude becoming to smart for ghumans??mrmm yess saar i installing claude now!tunak tunak tun tunak tunak tun da da da
>>107828224Please understand that this "human" offloaded his thinking into an AI as clearly shown by the fact that the tweet was written by an AI and not himself.
>>107822787>black and white anime girl pfp>copying kache>probably calls themselves tpotOpinion discarded.jpg
>>107822787>it's the economy that is brokenSchizo take.Supply and demand. New Avatar got like $390 million opening.It's all about who is willing and able to pay for your anime. That is who you target. That is who, therefore, determines the quality of your product.If the normie didn't pay for slop, the slop wouldn't be produced.
A huge number of niggers here are saying that AI code has security flaws, isn't maintainable, has memory problems, as if code written by humans doesn't have the same problem.So let's get to the facts: if both ways you end up with buggy code full of problems, why use humans when AI is practically free, won't file a lawsuit, won't go on strike, won't get sick, and will still work 24/7 without complaining?Currently, there is no reason to hire programmers, especially junior programmers and programmers without experience in general. Even mid-level programmers can be replaced now, leaving only senior programmers who know how to use AI.
>current AI model is the best!>next day>Hey guys this AI model now beats the benchmark!Literally wait 1-2 months and another model comes out to beat the current winner. It's all so predictable
>>107822787I hate this timeline. I wish AI (chatbots and generative AI) had never existed
East-Asians are already closer to the digital world than anyone else in the West due to living in a hivemind for centuries. They took over the West as the ones who mutated towards the future. At long as China and the rest of East-Asia keep Indians and jews out, they will take over the globe.
>there's no option to report twitter screenshot threads>reporting said posts won't make OP's PC explodeRAPEAPE/JAY, FIX THIS SHIT!
>>107829449>brand new technology being heavily explored by millions is improving everyday
>>107825562>These things literally don't have logical structure to their "thought processes"Neither do humans.You luddites would be utterly shocked and ashamed at how much ai research is teaching us about how our own minds work.
>>107829855Take all my data
>>107822787>ChinkseekNah
>>107822787>em dash
>>107831602I use it for improving my prompts
>>107822928>I don't even know how other white collar jobs can keep existing when they are already this good.Other white collar jobs have a benefit of models not being trained primarily for the core skill like Software Development is with coding models. There are compute, data and security bottlenecks & hurdles which will take more time to overcome. However it is inevitable and is gonna be less than a decade till it happens
>>107823022>The next step is to automate testing and Q/AThere are lot of heuristic aspects in QA which might be hard for AI to replicate. But I think quality won't matter much if buggy/glitchy software can be fixed on the fly by AI after users report it. So hard to predict what happens with QA
>>107823990>Third, AI of the future will have code that will be unmaintainble by humans, because they will be infinitely more smarter than humans at codingThis will happen when AI companies try to build end to end AI coding solutions with no humans in the loop
>>107825795Even without continuous learning, they are coming up with enough third party scaffolding that improves the AI model performance more than rawdogging it
>>107822787I like writing code, why the fuck would I use this shit?
>>107822787Threadly reminder that """AI""" doesn't exist and chatbots still can't do basic reasoning.Threadly reminder that 4chan no longer exists and this shithole is selling access rights to corporate bot farms.Threadly reminder that /g/ is moderated by literal troons who let AI advertisement threads (like this one) stay while deleting critical discussions.
>>107832451>scaffoldingLLM-coded corporate buzzword I forgot to add to the filter. Thanks.
>>107829172>why use humans when AI is practically freeIts not practically free yet, though it still beats paying for a junior dev
>>107825795>Continuous learning is a thing thats becoming a thingNo, it isn't. All you have is promises of meme hardware that's totally practical and absolutely hitting the market in Two More Weeks.
>>107832499half of /g/ must be filtered for you if you add all AI terms in filters
>>107829172>So let's get to the factsOk. Let's get to the facts. Which industry is deploying """AI""" to replace humans and how much did it profit from the switch? :^)
>>107832509>only half of the corporate botspam will be filtered if you filter itBetter than nothing.
>>107822787>expected>in the coming weeks>according to two people with direct knowledge of the plan
>some say
>>107832518>Which industry is deploying IA to replace humansJudaism>how much did it profitBigly
>>107830049>I’m retarded so everyone else must be tooAI really is the great equalizer. One way or another
>>107822787Claude has been beating deepseek recently though
>>107834078Claude has been the undisputed champion from Claude 2.1 onwards which is 2023. Just because normalfags are too retarded to use anything besides openai doesn't mean they were leading.
