I know a literal 80iq brown foid that has a coding job. Things like this make me doubt AI will ever replace coders to a relevant extent

>>25480745 It just means the people around her are even dumber regarding IT. But AI doesn't has to literally replace specific Coder X to have a impact. It makes everything more efficent, so 1 guy now can do the job of 5 or 10 people, which saturates the market. You can see the effects already now, new talent hiring has completly collapsed. Nobody hires new talent anymore. And the market being oversaturated will result in lower wages. If you top 10% or top 1% of coders, your wage will explode. If you just started, you are fucked.

Look this fool

AI now can write me a game with a genetic algorithm, where the agents all use language to actually think what their next step will be, and need to eat and spend energy while walking, and they can have offspring with other agents, and then you get 50/50 genetic mixture plus 10% mutation, so you can simulate evolution. And all this first try. Give any topic, and some scripter kiddy with AI will give in 24 hours better code than you could. Its literally better than 90% of paid programmers. And this is the worst it will ever be. Its only gonna improve

>>25480765 >If you just started, you are fucked. I'm going to kill myself.

>>25480765 Hiring has disappeared because all the 2020-2023 stimulus cash has dried up. Companies had buckets of money to throw away and now they don't.

>>25480814 What are you on about? The crash happened long before the stimulus checks. It has nothing to do with it. All tech companies did hire all new talent as a investment for the future, get them while they are young. Then AI came along, and they changed their strategy, and once the market was a bit over saturated, even more companies followed. They used to hire all new talent, now as new talent its almost impossible to get your foot in the door.

>>25480782 >AI can generate a shitty "game" that's just like a hundred other beginner programmer projects Woah, amazing.

>>25480847 We went from boilerplate and common algorithms to "Just shitty little games". It can code better than you already. Deal with it.

>>25480860 >boilerplate and common algorithms It's a combination of both. AI can do that too.

>>25480874 All code is a combination of something. Coders also don't generate their code out of a vacuum. Give any topic, and lets see who can write better code for it in 24 hours, you or AI. You just know the AI will write better code, and much more efficent and clean. Its already better than the majority of coders.

>>25480894 >Coders also don't generate their code out of a vacuum. Good programmers do.

>>25480894 Do you really think I'll spend 24 hours on some nonsense to prove you wrong? Are you brain damaged? Is this what AI does to people?

>>25480894 I hope AI keeps getting better and better and replaces as many anti-AI luddites as possible.

>>25480905 No, even a good coder doesn't generates code out of a vacuum. They were influenced and trained on other code. Its the same argument art fags make how AI never can make true art. Its absurd. This is the worst AI will ever be. AI will be able to make more and more complex things, and people will just dismiss it with "Pfffft its just combining already existing stuff". With art you forever can argue this point, but with code it won't matter. The only thing that will matter is the end result.

>>25480834 > stimulus checks Bruh, central banks everywhere pumped trillions of dollars into banks and businesses with 0% loans. Didn't you see all those videos posted here of retarded 22 year old girls with zero tech knowledge landing $200k jobs?

>>25480906 You know for a fact im right. Also im not even sure you are a coder to begin with. What coder didn't even play around a bit with AI and knows at what level it is right now. You are like 3 years behind.

Hahahaha

It doesnt tend to make mistakes with common algos and it can generate trivial boilerplate. Its basically a decent personal stack overflow machine. I've had it solve a few obvious problems for me that I couldn't solve because my brain was cooked.

>>25480918 Yes, but that happened all before coroner and such.

>>25480916 >What coder didn't even play around a bit with AI and knows at what level it is right now. Me. I think I prompted AI about 5 times when Dall-E came out to try to make some funny pictures and they all came out boring, then someone said ChatGPT programs well (don't even remember what version) and I prompted it once or twice to get it to output a program and that was it.

>>25480916 >No, even a good coder doesn't generates code out of a vacuum. ? You should absolutely have domain mastery by a certain point. You might occasionally have to google highly forgettable shit, which is where AI comes in useful.

>>25480925 Not him, but none of the LLMs I've used can handle more than 30 lines of code before doing retarded things. Even if you ask it to do some simple 3-4 line task, it will regularly generate terrible unusable code. They're more useful than search engines though, so I use them a lot.

>>25480969 This is about right. Its good for basic problems because you've forgotten something or are missing something. But it absolutely cannot replace an actualy engineer and it NEVER EVER will and if you think it will you're a fucking retard who does not understand how this retard prediction machine works.

>>25480782 How do they talk? LLM API?

>>25480782 >>25480782 >Its only gonna improve hold on there

>>25480925 The only thing I know about you is that you're completely detached from reality.

>>25480942 Thats not generating code out of a vacuum. You got trained on something. You didn't just sit in a cave with no input, and eventually you just code well.

>>25480976 >replace an actual engineer The average programmer does not deserve the title engineer.

>>25480969 >Not him, but none of the LLMs I've used can handle more than 30 lines of code before doing retarded things. Even if you ask it to do some simple 3-4 line task, it will regularly generate terrible unusable code. They're more useful than search engines though, so I use them a lot. I make gpt do 500-600 lines of code all the time, and at this point with o3 it works the majority of times at first try. Again, you are years behind.

>>25481011 True. Thankfully I was working in this sector for over a decade before the 'RETARD MACHINE' was added to all of our IDEs.

>>25480976 It will replace an engineer, and i know how that prediction machine works.

>>25480982 Whatever helps you sleep at night. The last 10 years i argue with bernds about AI, and every time i was right so far.

>>25481015 They cannot even produce prompts that long. Are you paying? LMAO.

>>25481015 They cannot even produce outputs that long. Are you paying? LMAO.

>>25481015 What? Lmao. The fuck it does. I have fed GPT4 component files, dependencies, and unit tests, and asked it to figure out why function X in the unit test was fagging out, giving it as much context as possible, and it just gives up and spits out fucking tech support questions, or completely randomised solutions to your problem. The second you challenge any of these LLMs with a complex, an actual complex enterprise level issue it rarely acts as anything more than a basic rubberduck. >>25481026 No, it fucking won't lmao.

>>25481040 Lol. Why even argue this if you have no clue? It takes 1 minutes to google it?

>>25481049 4 < 4o <<< o1-mini < o1 < o3-mini < o3-mini high <- We are here at the moment

>>25481065 I have GPT-4o mini and o3-mini at hand. Give me a prompt that produces a working 500 line program.

>>25481049 It will, no matter how much you cry and fight against it. You still are stuck thinking language models are just a glorified markov chain. Our brain is not magic, its also huge data sets and pattern recognition. It might be that we hit a wall with language models. But so far i don't see it. I don't claim i know for a fact just upscaling our current language models will create AGI that can replace coders. But they might. Or a combination of tech. While you claim 100% it never will replace a engineer. You literally can't know that and just talk out of your ass. You are as irrational as all these artfags claiming AI never ever can make real art. Its just pure cope. Soon we also get Byte Latent Transformer, the tokenizers improve at insane rates, we will get all kinds of hybrids and combinations. For all i know just having a few feedback loops with language models could alreaday get us first order logic/reasoning. But nope, you for a fact know AI never ever can replace a engineer. Im sorry, but any job that can be done from home will also be able to be done by an AI eventually.

>>25481081 "Scaffold a basic Springboot server to your most recent, up to date knowledge of Springboot. Provide me with a gradle file, a main file, and a component which exposes a /register POST endpoint which takes a username as an email, and a password. Define a service which provides validation against the username and password inputs, making sure the username is an email address, and the password is contains 1 special character, 1 number, and 1 upper case character. Write the appropriate SQL queries for insertion of these values. Finally, for what object methods you write; Create basic unit tests." It WILL fuck this up and it shouldn't be more than 500ish lines of code.

>>25481081 I don't give a shit what you have at hand. I use o3, and i get code that is 400-600 lines all the time. I think even deepseek can do 400 lines.

>>25481125 Nice cop-out.

>>25481133 How is that a copout? Are you retardet?

>>25481133 based btfo that germonkey >>25481140 stfu

>>25481143 Literally no arguments. > UHHH but my free version of chat gpt can't do it! So i disproven you! Ah ok.

>>25481149 I pay for chat gippity, or did until they announced I didn't need too. It failed, fails, and continues to fail with ever more context for enterprise level problems. The only way this gay shit is replacing actual developers en-masse is by burning all of the worlds oceans, building endless nuclear reactors to power it, draining the worlds coffers into Nvidia and making sure it never eats it own shit by scraping its own data off of the web. These machines will NEVER EVER reason the way a person does, it is INCAPABLE of it. You're weirdly dedicated to this being the truetrue of the future and you probably think Peter Thiel and Elon Musk are geniuses.

>>25481149 So you can't give a single example of your claims besides that screenshot of a badly made toy project?

>>25481174 I mean you keep doing this larp about your enterprise level problems, but i hope you are not actually retardet enough to just copy paste your enterprise level problems into a american AI, that sends all your data directly to america.

>>25481183 That toy project is what Chat gpt made for me. Like what else do you want?

>>25481184 Brother do you think I give a fuck about pasting UserManagementService.java into a Sam Altman's retard machine? What're they gonna do, steal my CompletableFuture<T>? You post like a fucking grad.

>>25481193 The onion asked you for a prompt, you didn't provide one. And nobody is impressed by a shitty toy project.

>>25481209 Sure, here is one prompt: Write a simple python game with a genetic algorithm. You have random blobs, little orbs, and they have random behavior you save as a genetic code, a string that defines their behavior. They move around on the map. There is also blobs of another color but less than them, that represent food. The blobs can also attack other blobs, and eat them, but they also can get killed in defense. The stats for their attack, and defense behavior is random too. 2 blobs can also get attracted and procreate, but for that it needs food. So only when they absorbed enough food, they can procreate with some other blob. The child blod is 50% genetic of one parents and 50% of the other parent, but there is always 10% mutation involved, so it for example could have entire different movement on the map, or better defense, or better attack, or is faster, or whatever. Its random. Thats the part of the genetic algorithm. Because the ones that can't survive in this environment get selected out. Also ad a function that if i click a blob with my mouse, it dies and leaves food behind. And if i click with my right mouse button i can leave food behind for them.

>>25481209 Why do you schizo pretend you are not the onion. And why are you so bitter? Im not trying to impress someone with code i didn't write myself. You sound exactly like artfags when people talk about AI art.

I made ans fractal with ChatGPT

>>25481222 There are at least 3 onions ITT.

>>25481235 Its over, artoids btfo forever, Elon Musk wins again.

>>25481174 >I pay for chat gippity, or did until they announced I didn't need too. What do you mean by that? > It failed, fails, and continues to fail with ever more context for enterprise level problems. You need to start fresh, AI still gets confused when the chat goes on for too long. > The only way this gay shit is replacing actual developers en-masse is by burning all of the worlds oceans, building endless nuclear reactors to power it, draining the worlds coffers into Nvidia and making sure it never eats it own shit by scraping its own data off of the web. Nobody claims current language models can replace developers or engineers. They can't even do first order logic yet, or at least the tokenizers can't. But its time to see the writing on the wall. We went in 3 years from AI not being able to reliable tell you how many legs a dog has, to now being able to explain you in detail why a dog evolved 4 legs.

Not too worried about llms replacing us codemonkeys. Average codebase contains thousands of pitfalls. You usually know about them after fucking around for a few hours, then a programmer who is older than the company tells you the reason something implemented this way. Almost every company will have some sort of inner-company lib to handle rare stuff because 10 years ago a *popular lib* didn't exist. A programmer would need to contact other people while debugging his shit, call some services from the other team, and try to not break anything during this process if he has any elevated rights in this shit. Of course that's because you need to write good code in the first place and then LLM will work but not even they output really good and thoughtout code. Let alone a typical codemonkey like me.

>>25481217 Even deepseek puts out 160 lines, and it worked first try. Can't wait to test grok, its apparently much better than chat gpt for this kind of stuff.

>>25481253 >Nobody claims current language models can replace developers or engineers. There is a literal retard ITT claiming it will do just that. >You need to start fresh, AI still gets confused when the chat goes on for too long. The moron machine cant understand context. How much more compute do we need to fucking feed it before it does? The best models around just melt the fuck down when provided more and more context. You can ask these moron engines to start from scratch over and over again, and use baby idiot googoogaagaa language to describe your complex problem over and over and over again and it simply cannot reason beyond its boundaries because it DOES NOT REASON. As I said, again and again, it cannot and will not replace reasoning engineers. It is at best a pocket stack overflow which comfortably acts as a rubber-duck when your brain is cooked from staring at your IDE for 6 hours, but only on the MOST BASIC problems.

He actually deleted his message lol.

>>25481279 >There is a literal retard ITT claiming it will do just that. Cool, then you can obviously quote me saiyng that, right? Im sorry if you are emotionally invested, but no, i didn't say current models can replace engineers. I know the limitations of this tech very well.

>>25481267 So it's not even a 500 LOC problem. >game with a genetic algorithm Jesus christ, who comes up with this shit? Is this some kind of standard benchmark for AIs? Have you ever even tried making it generate useful code?

Also it's been three years of "muh task solving in a second" and all of those tasks were: 1) solving tasks from code jams (the needed code is already known and is laying on the github) 2) creating simple games that no one would play anyway (students shit them out at production levels, search for snake or tic tac toe) 3) fixing a bug by placing a bandaid from first link from google (thanks could do that on my own)

>>25481286 >While you claim 100% it never will replace a engineer. You literally can't know that and just talk out of your ass. You are as irrational as all these artfags claiming AI never ever can make real art. Its just pure cope. Soon we also get Byte Latent Transformer, the tokenizers improve at insane rates, we will get all kinds of hybrids and combinations. For all i know just having a few feedback loops with language models could alreaday get us first order logic/reasoning. >While you claim 100% it never will replace a engineer. You literally can't know that and just talk out of your ass. it will never 100% replace an engineer. it will NEVER 100% replace an engineer.

>>25481296 >Jesus christ, who comes up with this shit? Is this some kind of standard benchmark for AIs? No, its just something i like to watch? I liked boxcar2d, so i make my own little algorithms with ai, to see them little creatures evolve. Why are you acting like such a bitter faggot? I don't care about benchmarking AIs, or solving world hunger with it.

>>25481302 Its literally a glorified stack overflow machine, it replaces a half hour of googling you highly specific retard problem. Which is useful, but its like saying 'Well if I just compile endless code samples from stack I will create software.'

>>25481308 So you shifted the goalpost again. First you claim i said current models already can replace a engineer. While i don't argue that. I say eventually AI can replace engineers. Why are you so dishonest? Are you drunk?

>>25481326 Guess what big man, the core logic and "reasoning" of these systems will never replace an actual engineer, no matter how much compute you throw at it.

What skills do I need to make sure I'm the one entering the prompts and not the one getting replaced by the prompt?

>>25481337 Tokenizers can't even do first order logic. But im not sure what basis you have to assume AI never ever will be able to do first order logic. I mean im kinda tired of arguing this. In 3-4 years we will see who is right. But i doubt you will be still here, and if you are, you won't admit you were wrong.

>>25481318 I'm acting like this because you're being an obnoxious bitch, who first claimed AI will replace all programmers and insulted everyone who disagrees with you and now are starting to backtrack and act like the victim. Do you even read your own posts?

>>25481356 Why do you call transforemers tokenizers? You do know there are some language models that don't tokenize their inputs and still work.

>>25481357 Where did i backtrack? Im very accurate in my speech, and i didn't backtrack on anything. The claim was AI will never replace a software engineer. Its obvious bullshit. Im not saying it will replace every single software engineer. The top 1-10% of software engineers will probably stay relevant. But it will replace some. No matter how much you cope about it.

>>25481365 Yeah for example Byte Latent Transformer. But its not like they are out yet for the public, or cheap to run. For now you need to tokenize somehow. Or what model is out there that doesn't tokenizes? Send me link.

>>25481355 How good are you at sucking dick?

>>25481356 Fine to put a peg in this, no problem at all. One of the things that AI stans forget, beyond the simple reasoning is they autistically reduce the process of development as a whole to a non-human game of logical interactions. This is a deranged field of thinking which clearly has no actual interaction with reality. The process of developing software and solutions to customer and client problems is inherently sloppy, and people-driven. This obnoxious, and inherently wrong need to attempt to assign behavioral needs to a system as complex as business relationships to algorithmic outcomes is pure autistic hubris. Engineers will exist, continue to exist, and will always be needed to define, implement, and architect the parameters of necessity of human interaction. An AI will not, and cannot, in their current form, replace these interactions. Corpos heavily investing in AI are actually kinda rolling it back a bit because the KPI outcomes are poor when you hyper-invest. The autistic utilitatian future, sadly, will not come to pass. You were cooked before you went into the oven, bro. rip.

>>25481374 You have yet to demonstrate the AI generating something that isn't a trivial a toy project glued together from common code examples you can find on the internet. Which is what you essentially claimed it can do when you made this thread.

>>25481394 You can keep repeating it, but we can just agree to disagree. Its data sets and pattern recognition. If a engineer can figure out thru text and voice over the internet what the customer wants, then a pattern recognition machine eventually will be able to do it too. For some reason you just want it to be true that Task X can't be done by a machine ever, but you will be wrong. Just how anti AI people were wrong for the past 5 years about every single thing. You can keep coping and laugh at AI generating 6 fingers or not being able to count how many letters are in a word, but as we go, every week AI can do more and more. And the claim is not that AI will replace every single engineer. AI will not replace every single engineer, every single artist, every single coder. But it will replace some, and will saturate the market. For a job where you needed a team, you will be able to do it with less.

>>25481416 Doesn't seems hard at all to just filter the shit code out.

>>25481384 Thought BLT still tokenized but in a different way.

>>25481414 Were you the one that claimed it can just write 30 lines of code?

>>25481445 No.

Onion coping because his keyboard monkey job is getting replaced.

Ai can do in 20 seconds for what bernd needs 2 weeks.