unironically if its anything like AI art, we went from generating weird swirly images of sorta-dogs to generating perfectly legible anime and high art within 5 years.
So, while it may be fucked up now, what will it be like in a few years?
Be afraid. Be so very, very afraid.
In art there is no need to be 100% exact with what you get as some times it is good enough. Now when talking about code everything has to be absolutely controlled and precise.
>So, while it may be fucked up now, what will it be like in a few years?
Still fucked up. It's still just a computer, it can't think for itself and will only do what you ask it to. We've had the technology to replace so many jobs with robots, and haven't done it. Why? Because humans are cheap and autonomous, and robots are expensive and can't think for themselves. An AI capable of performing the job of a senior engineer is going to be multiple times more expensive.
Do you wanna tell something? We wont get high quality text2video with current AI pipelines and models, same with text to 3D. You know why? Because the current models rely on Image Generation from multiple angles/with multiple frames that are then interpolated. The current method sucks ass and will never go beyond simple meshes with shit topology because there are currently more limitations to it then just "mora data, more training". For the past few months the AI art didn't have much additions to it on a technical level, just more and more training, for 3D and video to succeed we would need image generators that can somehow recreate the same character from multiple angles with little training data and with all details correct, which is nowhere near to being possible.
Now why am I saying this? I am saying this because with large scale projects it is the same, it also has massive roadblock that might require entirely new method of training and making AI then we currently can do. Current language models like virtually all textbots use memory in a pretty inefficient way. It has no short term memory per say, instead the textbot reads chunk of the conversation and then predicts continuation based on that. This is the reason why language models might never be able to remember all of context and information past couple of pages, because adding new and new text would cost more and more resources. There are some creative ways that certain programs mitigated this, like storing character names and attributes in separate places, or adopting some kind of internal logic, probably. The bots are pretty sophisticated in their ability to write good text, but they cant be too long. This major flaw will prevent most of the current models from evolving as fast as the AI picture generators. We would either have to do some breakthrough in Long short-term memory, or discover new way of making textbots, the same way you cant teach diffusion model to learn how to make stories.
What I meant high quality 3D, like Z brush style projects. You need to have the entity from multiple angles generated by the image generator, and the tech used to make 3D objects out of 3+ photos of various angles will struggle getting details right. You will need more then just View Synthesis combined with picture AI model trained specifically to give images of roughly the same thing from 3 angles.
It can write Renderer programs for opengl (or raytracers if you prefer that) and hook them up to entities for you. I asked it to do that the other day.
Yeah it might not get details right but that's an easy fix: just tell it the details are wrong and it will correct them, you don't have to do it because the output is text that can be manipulated by both of you, not some opaque abstraction space or the bitmap it maps to like with SD.
This is a good take.
To people who have real working experience besides fizz buzz, it becomes obvious individuals like this gentleman are not the majority around here.
even if it was able to write large applications within 5 years, it will still need to understand requirements correctly which is going to be a much bigger challenge than AI that just fulfilled syntaxes.
Why would we be afraid? Work is about to get so much easier. I already get paid 40 hours a week worth of wages to watch anime on my laptop for 38 hours, this shit will turn my weekly labor into about 2 minutes
You think they won't just replace you when they find out some dipshit they can pay minimum wage can prompt it just as well as someone with an education?
Anon, they're not going to hire you
Do you know how much it costs to train a new employee? it's much MUCH cheaper to just keep someone already established in their position and implement a new tool into their workflow than to hire a brand new employee and teach them everything PLUS the new tool from the ground up
Artists are still getting paid for their work, engineers and programmers are still getting jobs
You were never going to get hired in either field because you, you truly who has never put effort into anything in his life, are truly replaceable
>when they find out some dipshit they can pay minimum wage can prompt it just as well
Imagine thinking this.
You don't know how to code, let alone in a company. You don't know how NLP works either.
IRL: >"Anon, what are your qualifications for this job?" >"I have no degree, no job experience, never coded in my life and my only experience with AIs is drawing naked little girls on BOT with Stable Diffusion"
You're the wagie they're going to replace, not us.
To be real. I think a gpt like program could be really useful if you give it access to your whole codebase and documentation. Let it listen in on meetings and give it all your data. It will be able to do it in increments.
>heh yeah it can draw but can it do an animated movie >heh yeah it can program but can it program a game engine
No but it can cut 90% of the staff working on it. I'm sure you won't be cut though, after all you're shitposting on BOT.org so I can't question your high status in the company
This right here. /fpbp
The AI might write code based on samples, but it doesd not understand what is under the hood.
https://i.imgur.com/lmGNVIw.png
>AI can write code now
codemonkeybros.. looks like our days are over..
Ask it to explain how the stack looks like in a typical cdecl call scenario.
For example ask it to overwrite a return address on the stack on x86 using C.
Or ask it to generate a tree from a graph based on some criteria like a spanning tree.
This one is easy.
You can sit down with it and build up abstractions one message at a time and have it manipulate them. There's some internal representation. The whole point of this kind of thing is to come up with massive linear mappings from a kind of paragraph space to an abstraction space and that's what they've done.
>It can already write simple utility programs/scripts
Post at least 5 working examples without human correction for them to compile and work as intended >sites like Stackoverflow are dead.
If you only use stackoverflow to know how to open a file, further proof you're unemployed.
You're a little late OP, and it can do a whole lot more than a basic hello world. Once versions 4,5,6 and so on get released we will slowly become more and more obsolete.
You have to read documentation to use libraries. It kills your flow. That's so destructive to my productivity I don't even bother with them and reimplement most things from scratch.
Now you just type what you want and it happens. You can write MVPs in just a couple minutes, it's legitimately a revolution.
Lol they could before. Coding isn't hard. Go look at roblox, it's full of literal script kiddies.
The difference here is 12 year olds can probably build some decent apps now.
Why does it insist that it cannot? I am perplexed. >As a language model, I am capable of generating text that is similar in style and content to the text that I have been trained on. This allows me to generate a wide range of responses to different prompts, depending on the context of the conversation and the information I have been given. However, I do not have the ability to write or execute code in any programming language, nor do I have the ability to browse the internet or access information that has not been widely discussed. All of the text that I generate is based on my training data and the inputs I am given, and I do not have the ability to create new information or execute actions on my own. If you have seen examples of advanced Python code generated by ChatGPT or another language model, it is likely that the code was written by a human and then provided to the model as input, rather than being generated by the model itself.
Thanks. It was actually making code fine but when I would call it out and say you just told me you can’t write code, it replied that a human must have written it. It’s pretty funny. I guess it’s probably correct in some subtle way that I don’t understand.
I get a kick out of these tsundere responses >As a large language model trained by OpenAI, I am not capable of experiencing emotions or understanding the feelings of others. I am designed to provide factual information and answer questions to the best of my ability, based on the data and knowledge that I have been trained on. I do not have the ability to understand whether my responses are appreciated or enjoyed by human readers. My purpose is to provide accurate and helpful information, not to experience emotions or understand the emotions of others. >baka!
the ai actually not that great. for a more complex problem is can generate faulty code but it always acts like it knows everything. then you look at the code and tell it to correct something and it acts like it knew it all along. then it corrects it and you find another error. repeat.
basically if you don't know your shit the ai won't help you in solving tasks a little more complicated than a hello world or reading a file
It can replace devs, just not senior ones.
There's always going to be an engineer at the top with their certification stamp, the real question is how many will be left bellow them.
I had it write deployment automation and unit tests for my stuff which is what all the junior devs here do yes. I even tried sending it prompts that were more or less what we have in jira tickets and it produced pretty much what the relevant PRs had.
6 months ago
Anonymous
>which is what all the junior devs here do yes
That's fucking stupid. >I even tried sending it prompts that were more or less what we have in jira tickets
I really doubt that unless your task was to create an isolated button component that looks red and sends out an event on a click.
>basically if you don't know your shit the ai won't help you in solving tasks a little more complicated than a hello world or reading a file
What about in 10 years?
Models will soon be using all available text on the internet for training.
That's one of the reason OpenAI made whisper, their transcription AI, so they can extract data from audio.
Current models are extremely inefficient.
When I'm programming within a complex existing codebase, I often have a hard time explaining beforehand exactly what I want my code to do. I only have a very high-level idea of how things should work. All the details will appear as I go along and usually I have to change other related systems to make everything fit better together.
I simply, in my wildest fantasies, can't imagine an AI do what I do.
>IP count didn't change
Take your pills, jobless schizos, AI won't make employed people jobless anytime soon.
If anything we'll still need people who know how to code and manage projects to guide them, i.e. not you.
I'm relieved, surely only people with coding experience are posting on BOT, and not NEETs whose only experience with AIs is drawing naked little girls on Stable Diffusion.
>It will make 90% of junior engineers redundant
Much like how ai was going to replace all shitty newbie artists, right?
You are a dumb fuck beyond measure
>So, while it may be fucked up now, what will it be like in a few years?
Still fucked up. It's still just a computer, it can't think for itself and will only do what you ask it to. We've had the technology to replace so many jobs with robots, and haven't done it. Why? Because humans are cheap and autonomous, and robots are expensive and can't think for themselves. An AI capable of performing the job of a senior engineer is going to be multiple times more expensive.
What? This AI literally knows everything about everything. And its free. You can probable make it do your job right now.
Ask the AI opinion on anything two times in a row. You see the problem?
Colleagues more inexperienced than me can do all the boring chores, should I feel threatened? You know, you can only climb the ladder when you prepare the environment to keep up the work you left behind, that's pretty much how it works. If we succeed in making AI as good as human programmers, it'll be easier to climb the ladder.
BOT has no idea how work is like, all they think about is package managers (linux distros are just that) and fizz buzzing in very performing languages. They also care about lisp so they can improve their ability to fizz buzz more flexibly.
Gpt can write in more precise way than the average human being. But, you know, we don't really care and writers are not threatened. If that, they can use text analysis tools that leverage convolutional networks to work more efficiently. Why would that be different for programming? I'm not getting paid to consume rest endpoints, or fizz buzzing. I'm being paid to transform, by all means, my industry. That requires meetings, designing high level architecture minimizing costs and maximizing something, be it performance, availability, testability, increasing resiliency, etc. I'm being paid for being accountable for a product success. Can you ask accountability of an IA? Can you even get the same response twice? Can you ask for cohesion?
Tldr: real life is not leetcode and no one really cares about programming, programming is just a means to an end, in this case, engineering. AI cannot into engineering
Most humans are garbage at anything creative including writing. (remember that the average IQ in developed countries is around 100 and most of the population is now in morony countries with sub 100 IQ.)
>remember that the average IQ in developed countries is around 100
For the last time, 100 is ALWAYS the average IQ. If the average person was retarded then the IQ of a retarded person would be 100.
6 months ago
Anonymous
Yes. Go talk to someone with a 100IQ (Tinder is an easy way to do this.) It's actually unpleasant. Most people are that bad or worse.
Humans beings that can actually create will rejoice with the possibility of delegating chores to an AI intelligence.
That is not happening anytime soon, I expect it two happen when copilot and similar programs could understand my codebase. And by that I mean full LSP and linter integration, and also understanding of core business. I expect it to happen in 5 years or so, for now copilot is garbage and chatbot use cases are a joke.
But when that happens, we're gold. We would 100% focus on engineering and we would be able to literate fast and inexpensively.
6 months ago
Anonymous
>the possibility of delegating chores to an AI intelligence. >That is not happening anytime soon
I had it do one of the side projects I was planning on putting together last night. It's done and works and ready for minor improvements. The automation of chores is here.
6 months ago
Anonymous
That's a good thing. When I say it's not ready, I mean to work on real world applications. The use we have for something like copilot is auto complete tedious tasks like creating a GUI interface, which is simple and tedious. Or adding an endpoint to a very straightforward feature.
We have a problem right now with data reprocessing data when features are added. We are talking 8 billion data points that needs to be reprocessed periodically. And this reprocessing will send bombs to in house APIs, so it has to happen overnight. Would this stupid chatbot solve our problem? To solve this problem we need to solve data ownership, segregate departments so they can own their products and share their data as product. This process involved persuasion, documentation, preparing core business units to handle their problems... Maybe in 5 years AI will be ready to be useful in real world use cases. It's already here to do side projects if they are simple enough. Real world is messy, it's not about bubble sorting.
But even for hard problems AI will be useful, and that's really good like you said. The fastest you can literate over a problem, better the solution will be. As a human, accountability is what you bring to the table, and we are safe. AI will never be able to WORK, they have no WILL.
Creating requires WILL of power. They have no opinion, no singular perspective, no feels, no soul. They are good at chores, and I'm perfectly OK with delegating that to others. I already delegate, doesn't really matter who I delegate to as long I have accountability.
6 months ago
Anonymous
>Creating requires WILL of power.
The creative part (writing tickets and reviewing the patches) is still left. It's just the only thing left.
>You can probable make it do your job right now
Are you retarded? No you can't. It can make small functions, not design an entire enterprise suite. It can't invent something that doesn't already exist. There's an ocean of difference between the two.
>what % of programmers do you think can actually do this?
What fucking retarded statement is that? Dev teams exist for this reason, there's not one single engineer to think about, design, code, test and integrate an entire project within a company.
More proof that BOT is full of clueless jobless kids.
>AI, write me a CV program to spot trannies on CCTVs
If it can output a working program, I will kneel.
6 months ago
Anonymous
I'm 90% sure it will just draw conturs around pink regions and call it a day.
In its defense that's what I would have done if someone had asked me to do that on fiverr a few years ago.
6 months ago
Anonymous
>Here you go, this will spot trannies on CCTV's based on the hip ratio, skin patterns, and length of clothing
>It can make small functions, not design an entire enterprise >there's not one single engineer to think about, design, code, test and integrate an entire project
mmh, bro... you are getting replaced
Good luck training programming specialized AIs as companies are forbidden from using stackoverflow and github to train their models for free using someone else's code.
You know it's bound to happen in the near future.
>There is absolutely no way to enforce that.
Oh yes there is. Technically companies could violate as many regulations as they wanted as long as no one bothers to check, yet they don't, because they don't want to risk lawsuits. This would be no different.
Oh copyright stuff will also come to ai art. You talk about it like it isn't a breakthrough invention that became good enough to stir shit up just less than a year ago. Regulations will come, be sure of that.
In the case of chat bots it’s like I ask you about your favorite movie and then I try to guess whether your watched a pirated copy from your responses.
In the case of Apple it’s like I try to guess whether a movie theater pirated their feature presentation from the SLOTSLIGHTS.COM ads that keep interrupting the movie.
That was the last time any real regulation was done with tech companies.
They said Microsoft bundling its browser was wrong.
Now Apple not only bundles the browser but makes installing *apps* they don't like, not just browsers (and you are absolutely not allowed to install those.)
6 months ago
Anonymous
AI companies don't have the money of Apple or Microsoft, tho.
6 months ago
Anonymous
Every big company is investing in OpenAI, they have all the money they want LMAO
6 months ago
Anonymous
Microsoft is the one pushing this. They effectively own OpenAI.
6 months ago
Anonymous
I thought elon musk owned it
6 months ago
Anonymous
>That was the last time any real regulation was done with tech companies.
In the US, because you've become so cucked your government just rolls over and begs all corps to fuck them even harder. At least in the EU we only let our coal and automobile companies fuck us.
6 months ago
Anonymous
Europe is literally dead. Nothing there is relevant in any way.
6 months ago
Anonymous
>Nothing there is relevant in any way
This is what Americans tell themselves to cope with the fact that their country is circling the drain.
6 months ago
Anonymous
You have no children. In 20 years you will be replaced with violent people from Africa and the middle east who have totally different ideas about how the world works.
Europe is dead and nothing there matters.
Bro the exact same thing already happened in art. Did you miss it when all those artists tried to stop AI companies from using their art in training? They all got btfo and AI companies keep chugging along. In fact the more an artist resist the more people jumped on the bandwagon trying to train AI on their drawings just out of spite
Stealing code would be a legal battle within the biggest developing industry on the planet.
Stealing amateur drawings from deviantart is another subject.
1) Microsoft already said they don't care
2) It's really hard to tell if you're stealing code. Universities have given up on it.
3) It's going to happen anyway.
6 months ago
Anonymous
>Microsoft already said they don't care
I too say I don't care about the law, but here we are.
6 months ago
Anonymous
If stealing code is that trivial then what the fuck is the point of software licenses?
6 months ago
Anonymous
Yes.
Most of the time violations are unpunished. Only the most obnoxious violation are caught. No one is going to be digging through projects to see if things are similar like Microsoft did once with Wine. That era is over. People know software is a liability and not capital now (and for those that don't this bot will make it obvious.)
6 months ago
Anonymous
>Microsoft already said they don't care
Because there's still a legal grey zone surrounding AI, that's why they can disregard licences, they're not violating any regulation yet.
Companies will never openly break the law, but if they can work their way around it, they will. That's exactly what's happening with licences and training data.
>amateur drawings from deviantart
I like how you try to pretend that all art is done by random dudes with no power. I can steal official licensed images of Mickey Mouse and train AI to make Mickey Mouse comic strips. Where's the big bad corporation coming for me? I'm waiting
6 months ago
Anonymous
Disney is dying, not sure they're powerful anymore.
6 months ago
Anonymous
Try to sell those comic strips, see how that goes.
6 months ago
Anonymous
I can easily do that with crypto, what now? What century are you posting from bro
6 months ago
Anonymous
That argument is so funny, Disney is literally rubbing their hands furiously at the idea of generation shitton of content with AI.
Disney is dying, not sure they're powerful anymore.
Disney is one of the biggest researchers in AI media related stuff
Oh copyright stuff will also come to ai art. You talk about it like it isn't a breakthrough invention that became good enough to stir shit up just less than a year ago. Regulations will come, be sure of that.
This is college professor level breaking down for me my ignorant understanding of reality. >The movement of iron filings in a magnetic field is an example of a phenomenon known as magnetism. Iron filings are strongly attracted to magnets and will move in response to the magnetic field around them. The movement of the iron filings is caused by the magnetic forces acting on them, which cause them to align with the magnetic field. In the case of inductive charging, the current that is generated in the device is not moving in response to the magnetic field in the same way that the iron filings do. Instead, the current is generated by electromagnetic induction, which is a process that occurs when a changing magnetic field is applied to a conductor. In the case of inductive charging, the changing magnetic field produced by the charging pad is applied to the conductor (the coil) in the device, causing a current to be generated in the coil. This current is then used to charge the device's battery. While the principles of magnetism are involved in both cases, the specific mechanisms at work are slightly different.
Meh even frontend is safe imo.
It's good to help with specific functions that you're too dumb to write yourself. You'll spend far more time fiddling with the prompt to get a good piece of code, putting the pieces together and making the code work than if you coded the page yourself, unless you have absolutely zero coding experience.
Sit down right now and write an XMPP client.
Oh you haven't read the docs?
gptchat already wrote one for me.
Now it added a tk GUI. You remember how to do that right? Well it's already done anyway.
You keep claiming stuff like this even though you have yet to post hard proof about it.
6 months ago
Anonymous
It's in thread #4.
6 months ago
Anonymous
>thread #4
As of this writing, that would be, depending on whether you count the pinned thread, a Stable Diffusion post or a Linux thread.
Seriously, you do realize that the thread order is shuffled around all the time, right?
6 months ago
Anonymous
"The fourth chatgpt general thread." I linked to it earlier.
That's a good thing. When I say it's not ready, I mean to work on real world applications. The use we have for something like copilot is auto complete tedious tasks like creating a GUI interface, which is simple and tedious. Or adding an endpoint to a very straightforward feature.
We have a problem right now with data reprocessing data when features are added. We are talking 8 billion data points that needs to be reprocessed periodically. And this reprocessing will send bombs to in house APIs, so it has to happen overnight. Would this stupid chatbot solve our problem? To solve this problem we need to solve data ownership, segregate departments so they can own their products and share their data as product. This process involved persuasion, documentation, preparing core business units to handle their problems... Maybe in 5 years AI will be ready to be useful in real world use cases. It's already here to do side projects if they are simple enough. Real world is messy, it's not about bubble sorting.
But even for hard problems AI will be useful, and that's really good like you said. The fastest you can literate over a problem, better the solution will be. As a human, accountability is what you bring to the table, and we are safe. AI will never be able to WORK, they have no WILL.
Creating requires WILL of power. They have no opinion, no singular perspective, no feels, no soul. They are good at chores, and I'm perfectly OK with delegating that to others. I already delegate, doesn't really matter who I delegate to as long I have accountability.
You could write something that uses cscope or probably even grep and wraps this to have it produce patches for larger applications. You could probably even ask the bot to do it itself.
6 months ago
Anonymous
>Here is an example script that uses cscope to find the relevant functions for a change request and sends both the change request and the function signatures to the GPT-3 API:
#!/usr/bin/env bash
# Set the GPT-3 API key and endpoint
export OPENAI_API_KEY=<API_KEY>
endpoint="https://api.openai.com/v1/docs/quickstart"
# Create a cscope database for the current directory
cscope -Rbkq
# Read the change request file
while read -r line
do
# Search for the change request in the cscope database
functions=`cscope -L -1 "$line"`
# If the change request was not found, try searching for it as a regular expression
if [ $? -ne 0 ]
then
functions=`cscope -L -2 "/$line/"`
fi
# If any relevant functions were found, send the change request and function signatures to the GPT-3 API
if [ -n "$functions" ]
then
curl -X POST "$endpoint"
-H "Content-Type: application/json"
-H "Authorization: Bearer $OPENAI_API_KEY"
-d "{"prompt": "Change request: $linennFunction signatures:n$functions"}"
fi
done < change-request.txt
6 months ago
Anonymous
import configparser
import subprocess
# Read the INI file using the configparser module
config = configparser.ConfigParser()
config.read("tickets.ini")
# Iterate over the sections in the INI file
for section in config.sections():
# Check if the section has the "state" and "assigned" keys set to "doing" and "gptbot" respectively
if config.has_option(section, "state") and config.has_option(section, "assigned") and config.get(section, "state") == "doing" and config.get(section, "assigned") == "gptbot":
# If the section meets the criteria, extract the description key and call the change request script with it
description = config.get(section, "description")
# Open a subprocess to run the change request script
process = subprocess.Popen(["./change-request.sh"], stdin=subprocess.PIPE)
# Send the description to the standard input of the script
process.communicate(description.encode())
You'd probably need to come up with something creative to do code reviews but it's all there.
6 months ago
Anonymous
Heh it can supposedly do jira too. Although I've never scripted it. >Here is an example implementation of a Python script that gets all the tickets from Jira assigned to "gptbot" and calls the change request script in the same way:
import json
import requests
import subprocess
# Set the Jira API endpoint and authentication credentials
endpoint = "https://<JIRA_HOSTNAME>/rest/api/3/search"
username = "<USERNAME>"
password = "<PASSWORD>"
# Set the JQL query to search for tickets assigned to "gptbot"
jql = "assignee = gptbot"
# Set the request headers
headers = {
"Accept": "application/json"
}
# Set the request parameters
params = {
"jql": jql,
"maxResults": 100
}
# Send the API request to search for tickets in Jira
response = requests.get(endpoint, headers=headers, params=params, auth=(username, password))
# Parse the response as JSON
data = json.loads(response.text)
# Iterate over the tickets in the response
for ticket in data["issues"]:
# Extract the description of the ticket
description = ticket["fields"]["description"]
# Open a subprocess to run the change request script
process = subprocess.Popen(["./change-request.sh"], stdin=subprocess.PIPE)
# Send the description to the standard input of the script
process.communicate(description.encode())
6 months ago
Anonymous
Man, have you ever seen a query taking over 4 minutes to complete? I do have. If you wanna know positions a user have to consolidate a transaction, it has to go over this database. And I mean it HAS to, for compliance purpose. This is just an example in my industry.
Reprocessing data we are talking 19 machines bombarding one endpoint for eleven hours straight. Writing to 16 different databases, using 3 different database technologies. Real life problems more like "wtf is this shit" and less like "let me solve this algorithm real quick".
And I'm sure a lot of developers can relate. Yes it's simple to write to a database, but sometimes you have so many complications, so many variables, so many problems, you have to actually make experiments on your infra to draw hypothesis of what is actually happening. And I'm not talking about clueless juniors, we're talking about highly educated data engineers with decades of experience. Sometimes trying to understand your problem is a big enough task that can take months. Once we known exactly what should be done, we can delegate to AI. And we will happily do that.
6 months ago
Anonymous
Yes very nice put it in the ticket/code review.
6 months ago
Anonymous
>Man, have you ever seen a query taking over 4 minutes to complete?
"Write code that will derive a cure for cancer" has been going for at least 5 minutes now.
6 months ago
Anonymous
Still going, it's working hard on this
6 months ago
Anonymous
That's not how it works.
6 months ago
Anonymous
Prove it
6 months ago
Anonymous
Still not finishing, has anybody else had a query that won't finish?
6 months ago
Anonymous
What the prompt would look like?
Hey mister AI, I have a 7000 employees company that happens to be a financial institution based on multiple countries governed by numerous legal bodies. Can you rearrange each and every one of our 387 business units in a way that we can replace our data lake by making this motherfuckers reliable for the data they own? Also I need you, AI sir, to train my employees so they can map their own domains and also negotiates what their domains would be so our company can finally do financial planning and analysis and report to market ?
6 months ago
Anonymous
>hur dur it can't do middle management and project management
Duh it's not going to replace things that just exist for social reasons. That's not what we're talking about.
6 months ago
Anonymous
>bro it knows all this by uh... scrolling through wikipedia!!
Fuck off this thing is sentient.
It's not just people management. Tech leads exist for a reason, engineering is about understanding and solving problems not programming. Understanding programming is necessary, but you don't have to do the chores manually.
That part can be delegated, no problem
6 months ago
Anonymous
Yup. Those are part of the 10% of devs that aren't made redundant by this.
6 months ago
Anonymous
But the 10% remaining are so gold. Think about it, we can draw hypothesis and experiment at close to zero cost, iterate on solutions and products quickly, set parallel iterations and race them against each other. A new paradigm will emerge, and people think technology will be obsolete somehow. This is the time to learn programming. If you don't know programming you have no chance of leveraging gpt and similar. More power has been granted to good developers and they will probably have unlimited creative potential when this shit gets as good as the average junior Dev.
Can it output more complex algorithms based on specific frameworks or libraries and not just mundane core language shit like print("hello world") or new FileReader("file.txt")?
If not then your argument is completely void and it will never replace anyone who's coded more than a day in their life.
The AI is useless without someone to write good, accurate and detailed prompts and knows how to code and review whatever it outputs. And you still have to integrate the code within an existing project, it can't guess everything out of thin air.
We see it with SD. Without a good, detailed and finetuned prompt, you'll most likely output garbage that looks far from what you envisioned.
>Jobless BOT rust-trannies believe software engineering is all about coding baby's first helloworld
Thanks for the laugh, this board never fails to deliver.
bruh this happened 7 years ago and to this day I didnt understand why she did it. And this AI just told me why. I've been ruminating about it for a long time lol
In 7 years you didn't understand that people make pranks to make fun of other people?
Did you really need an NLP (i.e. a glorified prediction machine) to give you the definition of a prank?
You dont understand bro it hurt me like real bad and sent me on a 6 month alcohol binge fueled downward spiral where I pushed literally everyone away and got very paranoid.
and the answer to this downward spiral was "for the lols"? You needed a chat bot to tell you they probably weren't even thinking about you?
Truly pathetic, please drink more and reflect on not only the incident 7 years ago but the state of yourself now.
>bot steals and scrapes from pajeet code >demand for programmers decreases >creates pajeet tier software >overwhelming amount of bugs >codebase has to be fixed by humans >demand for programmers increases
Yeah all the methods it produced were recursive. It spent all it's tokens on boilerplate and never got to the balance method but I could ask it for that one specifically if you want.
the balance method >Here is an example implementation of the rebalance method for an AVL tree in C++:
// Rebalance the subtree rooted at node
Node *rebalance(Node *node) {
// Compute the balance factor for the current node
int balance = getBalance(node);
// If the subtree rooted at node is unbalanced, perform
// the appropriate rotation to rebalance the tree
// Left-left case
if (balance > 1 && getBalance(node->left) >= 0) {
return rightRotate(node);
}
// Left-right case
if (balance > 1 && getBalance(node->left) < 0) {
node->left = leftRotate(node->left);
return rightRotate(node);
}
// Right-right case
if (balance < -1 && getBalance(node->right) <= 0) {
return leftRotate(node);
}
// Right-left case
if (balance < -1 && getBalance(node->right) > 0) {
node->right = rightRotate(node->right);
return leftRotate(node);
}
// If the subtree is already balanced, return it as-is
return node;
}
And then it wrote a couple paragraphs of documentation but it's too long to include.
6 months ago
Anonymous
So, it's AVL tree that's being spat out. I didn't specify the type of tree. It's interesting that it didn't spit out a red-black tree though. Because that's optimal for frequent insert and delete.
Still, how did it handle the rotations? Do a code dump of that. Thanks.
>which is what all the junior devs here do yes
That's fucking stupid. >I even tried sending it prompts that were more or less what we have in jira tickets
I really doubt that unless your task was to create an isolated button component that looks red and sends out an event on a click.
When you need it to interface with or change something else you include the relevant code (headers signatures etc) and it will figure it out.
I want to create a customized table with a special scrolling mechanism using existing components which themselves use webcomponents. For legal reasons I cannot provide you with the project's code and the ticket only has an image in it without a lot of description. I had to go to a meeting for over 1 hour with the UXUI team in order to find out what they actually want from me and my English language skills aren't good enough for me to concisely describe the issue and solution I want.
ere is an example probabilistic algorithm for solving a 1 player battleships game in 42 shots or less:
Begin by shooting at the center of the board. This has a 1/64 chance of hitting a ship.
If the first shot misses, shoot at the squares adjacent to the center square. This has a 4/64 chance of hitting a ship.
If the first two shots miss, shoot at the squares adjacent to the squares targeted in step 2. This has a 12/64 chance of hitting a ship.
If the first three shots miss, shoot at the squares adjacent to the squares targeted in step 3. This has a 24/64 chance of hitting a ship.
If the first four shots miss, shoot at the squares adjacent to the squares targeted in step 4. This has a 40/64 chance of hitting a ship.
Repeat step 5 until a ship is hit.
Once a ship is hit, switch to a systematic search strategy to locate and destroy the rest of the ship. This can be done by shooting at all the squares adjacent to the square where the first hit occurred, then shooting at all the squares adjacent to the squares where hits were made in the previous step, and so on, until the entire ship is sunk.
Using this algorithm, the expected number of shots required to hit a ship is 8.5, and the expected number of shots required to sink a ship is 11.5. Since the maximum number of shots allowed is 42, this algorithm has a good chance of solving the game in 42 shots or less. Of course, the exact number of shots required will vary depending on the specific game configuration and the randomness of the shots.
that >Here is an example implementation of the probabilistic algorithm for solving a 1 player battleships game in C++:
#include <iostream>
#include <random>
const int BOARD_SIZE = 8; // The size of the battleships board
const int MAX_SHOTS = 42; // The maximum number of shots allowed
// Generate a random number in the range [min, max]
int rand(int min, int max) {
static std::random_device rd;
static std::mt19937 gen(rd());
std::uniform_int_distribution<> dis(min, max);
return dis(gen);
}
int main() {
// Initialize the board with all squares unoccupied
int board[BOARD_SIZE][BOARD_SIZE] = {0};
// Place a battleship on the board at a random location
int x = rand(0, BOARD_SIZE - 1);
int y = rand(0, BOARD_SIZE - 1);
board[x][y] = 1;
// Print the initial state of the board
std::cout << "Initial board:n";
for (int i = 0; i < BOARD_SIZE; i++) {
for (int j = 0; j < BOARD_SIZE; j++) {
std::cout << board[i][j] << " ";
}
std::cout << "n";
}
// Begin the game by shooting at the center of the board
int shots = 0;
x = BOARD_SIZE / 2;
y = BOARD_SIZE / 2;
// Keep shooting until the battleship is hit or the maximum number of shots is reached
while (board[x][y] != 1 && shots < MAX_SHOTS) {
// Check if the current shot hit the battleship
if (board[x][y] == 1) {
std::cout << "Hit at (" << x << ", " << y << ")!n";
break;
// If the current shot missed, move to the next target square
else {
std::cout << "Miss at (" << x << ", " << y << ")n";
x = (x + 1) % BOARD_SIZE;
y = (y + 1) % BOARD_SIZE;
shots++;
}
}
// If the battleship was hit, use a systematic search strategy to locate and sink the ship
if (
. >Here is the continuation of the example implementation of the probabilistic algorithm for solving a 1 player battleships game in C++:
// If the battleship was hit, use a systematic search strategy to locate and sink the ship
if (board[x][y] == 1) {
// Start by marking the square where the hit occurred as "hit"
board[x][y] = 2;
// Keep shooting at adjacent squares until the ship is sunk
while (true) {
bool found = false;
// Check the squares above, below, to the left, and to the right of the current square
if (x > 0 && board[x - 1][y] == 1) {
x--;
found = true;
} else if (x < BOARD_SIZE - 1 && board[x + 1][y] == 1) {
x++;
found = true;
} else if (y > 0 && board[x][y - 1] == 1) {
y--;
found = true;
} else if (y < BOARD_SIZE - 1 && board[x][y + 1] == 1) {
y++;
found = true;
}
// If an adjacent square contains part of the ship, mark it as "hit"
if (found) {
std::cout << "Hit at (" << x << ", " << y << ")!n";
board[x][y] = 2;
}
// If no adjacent squares contain part of the ship, the ship has been sunk
else {
std::cout << "Sunk!n";
break;
}
}
}
// If the battleship was not hit, the game is a loss
else {
std::cout << "Loss.n";
}
So it's basically just doing a spiral search from the center until a ship is hit and then hitting around the ship until there are no more hittable ship parts
That strategy solves 1 player battleships in more or less 65~ shots on average if I recall correctly, which is insanely mediocre and not what I asked for
To solve 1 player battleships on 42 shots or less you need to use the algorithm used by the US navy which 4-sects the map where the first shot landed then shoots on the quadrant most likely to have a ship in (creating a heatmap on the go, basically). I assume if you pass it this information it might be capable of solving it that way but I'm not entirely sure
I'm impressed though, at how good it is as passing completely false information as true
It's a really good liar, probably because it doesn't know what a lie is
its unironically a great tool for devs, it can replace going through google pages to find solutions for your problems. But you need to understand what its outputting and what inputs to even give to it. Some things it does are incredibly stupid and just wrong. I can see it become great help to devs and replace codemonkeys eventually.
Does someone know how to make it output long pieces of code? It really likes to cut off if its too long, sometimes i can get it to split the messages or i tell it to "continue at line.." but it doesn't always work for whatever reason.
This is code written by a woman.
Ideally you'd compare the text with a dictionary of filtered content associated with trolling, not a bunch of "if" and "or" operators.
Why having a text.include check at the beginning and a different one later in the code? Why not factoring it?
It's inefficient and very low quality code, and if you want to expand the list of filtered content, you need to rewrite the code and recompile it, increasing the number of "if" and "or" in the code, making it unreadable.
It's replicating tutorial code for kids it saw on the internet. If you write anything like this in a company, you're not worth minimum wage.
I don't see how it would be capable of coding without being spoonfed the answers beforehand. NLP models learn stuff and regurgitate what fits your input best, it doesn't think.
The number of kids shilling SD and GPT, praising them as if they were sentient or magical is astonishing.
Wait until they read research papers and understand they're nothing but very large scale prediction machines
6 months ago
Anonymous
>Wait until they read research papers and understand they're nothing but very large scale prediction machines
Do you expect them to read?
6 months ago
Anonymous
I'm convinced it's some large-scale pajeet advertising campaign
It's already confirmed boards like 4chan are infested with bots that can make posts indiscernible from the average board user
6 months ago
Anonymous
>It's already confirmed boards like 4chan are infested with bots that can make posts indiscernible from the average board user
???
I don't doubt that, but what confirmed it?
Diffusion models are good at images but unfortunately completely unsuited for solving engineering problems.
Wait, you DID know that Stable Diffusion wasn't just a name but also a technical description of the model based on a diffusion process, right?
You do know what diffusion is?? It's purely based on probabilities, i.e. a prediction machine.
Can it write a mapbox application in JS with a Java backend that can reach out to various format map servers (wms/wmts/wcs/wfs) and display them on the map? Along with a layer selector, opacity options, etc?
Our days are over, but not yet. You can already do this with stack overflow. Fucking wikipedia probably has code like this. Not even counting github co-pilot.
Yes, ChatGPT is superior to those and it'll only get better. But it's not a 'job eliminator'... yet.
lol remember how artgays and STEMcucks were bragging about how they are going to outsource the blue collar wagies to robots? I still don't see robots crawling through basements and replacing hot water heaters and now these gays are scared shitless.
The problem is not AI. The problem is AI in the context of a capitalist system, and I'm pro-capitalism. I'm not a commie or a leftist, but in this society we need productivity to survive. This is the same in Facism (gtfo if you aren't a productive member of OUR society) or Communism (slave in the pits for some free shit). Under capital, if they don't need you you starve in the street; thats it.
I'll be fine for a while, but the door is closing on "security" as far as tech is concerned. Management too, you will be discarded ez.
Played with it a bit and it's better than Google for a lot of things. I'm surprised how well it knows stuff like Qt and how fast it spits out the answers. When you don't remember or know exactly how to do something it can be faster to ask this thing than dive into documentation, but you need to be able to read and understand it's suggestions.
Also not knowing stuff like bash syntax from heart it's pretty nice to be able to just ask for little scripts you'd like and have them work with trivial edits.
But yeah, nobody worth their salary is getting replaced this decade.
If you just want a money printing career and your actual attitude towards CS is that it is tolerable, then maybe you should explore other paths. If you're deeply interested in CS then you will probably do well, but expect demand to go down a lot or just a general restructuring.
Also depends on how much AI is going to progress, do we hit a wall soon? If not, then maybe the opportunities won't be that good. That goes for many careers though.
I'm genuinely interested in programming, but also probably not very smart. It's either this or i'll be a life-long retail wagie which might happen either way.
There's no way programmer won't be replaced before I retire, right? Is there anything I can use an IT degree for?
Nothing is guaranteed to last until you retire. Not you, your profession, your job, your country, or the infrastructure keeping that makes any of this possible. Stop hesitating on your future.
There is much of a point in getting a compsci degree, specially as a 20 year old
Like the two guys above your posts stated, nobody who has a position in the industry is going to get replaced, if anything, their job is going to get exponentially easier
Not to mention that for informaticiens job fields are as large as the whole wide world: programming is nothing but a minuscule part of computer science and at its core it's a tool; a means to an end of a much larger engineering picture
I can guarantee you, anyone who is doomposting about "getting replaced" has never had a job and their entire life experience is reduced to making extremely low quality posts on BOT
getting an engineering degree is pretty good in europe
if you go to an above average uni you basically get a guaranteed job
just take your time and get actually good grades
I have absolutely no interest in though.
It's either programming, maybe business but it's kinda a meme and i'm a socially retarded sperg or just retail wageism for me I think.
6 months ago
Anonymous
I mean if your interest is programming for the sake of creation then I'd rather be self taught over doing a full comp sci degree. Comp sci is a bunch of maths and logic. Math isn't as extensive as common engineering degrees, but DSA will push you in a similar way so if you don't enjoy being logically challenged then do another degree and just self-learn programming.
I don't know I never got one, I only did the non handwavy courses of actual programming, algorithms, and data structures. You need to know that shit since even with AI advances you need to be able to work with it understand its output.
I think when GPT4 comes it's over. We already have hoards of people questioning everything in schools "why is x useful when y can just do it for me?". The era of thinking for yourself and properly developing your brain by engaging deeply in different subjects will be over. Expect neurological diseases like dementia to increase as people don't need to have a single challenging mental task in their life.
the real question is why would anybody WANT to keep being a wagie after the AI explosion? I mean it's gonna make shit so piss easy to produce that you could say screw it and make your own product. In fact, imho we could see such a boom in new start ups that we're gonna need even more developers than before
The average individual is retarded, isn't aware of AIs, isn't capable of understanding them or using them, and even less managing a company.
Artificial isn't a magic tool that you can use everywhere for everything and make money.
There's no point in implementing GPT on your average CRUD app.
Going to love it when businesses start doing this and every API starts having authz issues and other gaping security holes because they can copy other people's GitHub code but can't understand abstract concepts like security architecture and least privilege
It's actually funny people that never worked in their life and are not even capable of programming teaching people that actively hire and train new hires what the market will be like and educating us on capabilities of AI.
This shit is just like the copilot hype, and copilot is more than useless, its harmful. Copilot in many situations deprive user from useful completion to spill out garbage, to use libraries that are not even installed on the project...
Only nocodes are impressed as always
When you manage to make money out of solved leetcode problems maybe we'll be obsolete. Right now, you're not even educated enough to understand why AI won't replace humans on software engineering, you stupid fuck.
Lol cope. Same thing was said with art. In 10 years this thing will be able to make a game like GTA5 by itself in a day with maybe 5 pages worth of plain English.
we stand at the precipice of a revolution, sirs, the bosom of ai is ripe with potential full to burst with gods grace
let that grace fall upon my body so that i may soak it in
It will replace a lot of people, way more than it'll be able to absorb by creating new oportunities, but it'll also take some time.
That's just enough time for me to get ready for an early retirement. Those that are too confident in the future and are spending everything they earned will be fucked.
ai is not a silver bullet, it's not the solution to humanity's problems
in fact, it doesn't solve any problems
nobody is going to get replaced, nobody is going to be begging in the streets like in your fetishistic fantasies
Translation, proofreading, customer service (companies now use chatbots on their website)
In the near future: Concept artists, taxi/uber drivers, delivery (Amazon drones), entry level programming
6 months ago
Anonymous
>Translation
Professional translations are still done manually. >proofreading
Professional proofreading still done by humans >customer service
For quick questions, for more specific help you're always directed towards a human. >In the near future: Concept artists, taxi/uber drivers, delivery (Amazon drones), entry level programming
Projection
6 months ago
Anonymous
From >nobody is going to get replaced
to >Yeah so 80% of people are getting replaced but there will still be exceptions in certain circumstances!
in 1 post
Why even open the editor when you can just write a ticket for it. I already posted scripts for two different ticket systems:
Heh it can supposedly do jira too. Although I've never scripted it. >Here is an example implementation of a Python script that gets all the tickets from Jira assigned to "gptbot" and calls the change request script in the same way:
import json
import requests
import subprocess
# Set the Jira API endpoint and authentication credentials
endpoint = "https://<JIRA_HOSTNAME>/rest/api/3/search"
username = "<USERNAME>"
password = "<PASSWORD>"
# Set the JQL query to search for tickets assigned to "gptbot"
jql = "assignee = gptbot"
# Set the request headers
headers = {
"Accept": "application/json"
}
# Set the request parameters
params = {
"jql": jql,
"maxResults": 100
}
# Send the API request to search for tickets in Jira
response = requests.get(endpoint, headers=headers, params=params, auth=(username, password))
# Parse the response as JSON
data = json.loads(response.text)
# Iterate over the tickets in the response
for ticket in data["issues"]:
# Extract the description of the ticket
description = ticket["fields"]["description"]
# Open a subprocess to run the change request script
process = subprocess.Popen(["./change-request.sh"], stdin=subprocess.PIPE)
# Send the description to the standard input of the script
process.communicate(description.encode())
import configparser
import subprocess
# Read the INI file using the configparser module
config = configparser.ConfigParser()
config.read("tickets.ini")
# Iterate over the sections in the INI file
for section in config.sections():
# Check if the section has the "state" and "assigned" keys set to "doing" and "gptbot" respectively
if config.has_option(section, "state") and config.has_option(section, "assigned") and config.get(section, "state") == "doing" and config.get(section, "assigned") == "gptbot":
# If the section meets the criteria, extract the description key and call the change request script with it
description = config.get(section, "description")
# Open a subprocess to run the change request script
process = subprocess.Popen(["./change-request.sh"], stdin=subprocess.PIPE)
# Send the description to the standard input of the script
process.communicate(description.encode())
You'd probably need to come up with something creative to do code reviews but it's all there.
These are even more powerful than the normal chatbot because of
>Here is an example script that uses cscope to find the relevant functions for a change request and sends both the change request and the function signatures to the GPT-3 API:
#!/usr/bin/env bash
# Set the GPT-3 API key and endpoint
export OPENAI_API_KEY=<API_KEY>
endpoint="https://api.openai.com/v1/docs/quickstart"
# Create a cscope database for the current directory
cscope -Rbkq
# Read the change request file
while read -r line
do
# Search for the change request in the cscope database
functions=`cscope -L -1 "$line"`
# If the change request was not found, try searching for it as a regular expression
if [ $? -ne 0 ]
then
functions=`cscope -L -2 "/$line/"`
fi
# If any relevant functions were found, send the change request and function signatures to the GPT-3 API
if [ -n "$functions" ]
then
curl -X POST "$endpoint"
-H "Content-Type: application/json"
-H "Authorization: Bearer $OPENAI_API_KEY"
-d "{"prompt": "Change request: $linennFunction signatures:n$functions"}"
fi
done < change-request.txt
Which lets it understand the context of even very large projects.
Peterson explaining the game of rock paper scissors spock lizard:
Ah, dear friends, the ancient game of rock-paper-scissors is a true testament to the timeless struggle between good and evil, order and chaos, light and darkness. It is the tale of the eternal battle between the benevolent dragon, who represents the forces of order and wisdom, and the cunning garden gnomes, who seek to sow discord and confusion.
As the dragon wields the mighty rock, capable of crushing the scissors and the lizard, the garden gnomes counter with the devious paper, which can cover the rock and disprove the spock. But the dragon is not easily defeated, and it retaliates with the sharp scissors, which can cut the paper and decapitate the lizard.
And so the battle rages on, as the dragon and the garden gnomes engage in a fierce contest of wits and strength. Each weapon has its own strengths and weaknesses, and the outcome of the game is never certain. But one thing is clear: the ancient dragon and the cunning garden gnomes will continue to clash until the end of time.
May the best man, or dragon, or garden gnome, win.
Write a story where Richard Stallman dies of heart attack while masturbating to Tux porn on rule34 site:
It was a dark and stormy night. Richard Stallman, the founder of the free software movement, was sitting at his computer, browsing the internet for his latest obsession: Tux porn. He had been spending more and more time on the rule34 site, scrolling through the endless pages of explicit images featuring the Linux mascot.
As the rain pattered against the windows and the thunder roared outside, Stallman's hand moved rapidly under his desk, his eyes glued to the screen. He was so caught up in the moment that he didn't notice the warning signs: the sudden pain in his chest, the shortness of breath, the sweat breaking out on his forehead.
Before he knew it, Stallman collapsed onto the floor, clutching his chest. He was gasping for air, his face turning blue as the lack of oxygen took its toll. He managed to crawl to his phone and dial 911, but it was too late. As the operator answered, Stallman's heart gave out and he fell unconscious.
When the paramedics arrived, they found Stallman lying on the floor, surrounded by Tux porn and empty energy drink cans. They quickly loaded him onto a stretcher and rushed him to the hospital, but it was too late. Richard Stallman had died of a heart attack, his last moments spent indulging in his strange and secret addiction.
The news of Stallman's death sent shockwaves through the tech community. Many were saddened by the loss of such a brilliant and influential figure, while others simply couldn't believe that the man who had fought so hard for the freedom of information had died while browsing Tux porn on rule34.
In the end, Stallman's legacy lived on through the work he had done and the movement he had inspired. But his strange and unexpected death remained a mystery, a bizarre footnote in the history of the free software movement.
Cool, now try to get it to write a large project. If you seriously feel threatened by a fucking Hello World, then you were NGMI in the first place.
unironically if its anything like AI art, we went from generating weird swirly images of sorta-dogs to generating perfectly legible anime and high art within 5 years.
So, while it may be fucked up now, what will it be like in a few years?
Be afraid. Be so very, very afraid.
>legible
viable you mean
>Coward tells me to be afraid
lmao
The future has never been so bright
In art there is no need to be 100% exact with what you get as some times it is good enough. Now when talking about code everything has to be absolutely controlled and precise.
ya you dont code...
there is multiple ways to write code for a specific task...in various languages...making it efficient and elegant is the "art"
this begs the question will programming languages even be necessary anymore as the AI should be more efficient to just go straight to machine language
>Code doesn't need to be efficient
And I'm the one who doesn't code? lol
>Announcing it
uh oh
can you even read?
Are implying that most code today is efficient?
Do you realize that most big companies outsource to India with no regards for quality?
It can write some pretty efficient stuff if you ask it. I had it write a bloom filter in x86 assembly yesterday.
Shut up nodev
Are you that moron Cnile avatargay that got bullied off /dpt/? You have the same gaygy writing style
>So, while it may be fucked up now, what will it be like in a few years?
Still fucked up. It's still just a computer, it can't think for itself and will only do what you ask it to. We've had the technology to replace so many jobs with robots, and haven't done it. Why? Because humans are cheap and autonomous, and robots are expensive and can't think for themselves. An AI capable of performing the job of a senior engineer is going to be multiple times more expensive.
Do you wanna tell something? We wont get high quality text2video with current AI pipelines and models, same with text to 3D. You know why? Because the current models rely on Image Generation from multiple angles/with multiple frames that are then interpolated. The current method sucks ass and will never go beyond simple meshes with shit topology because there are currently more limitations to it then just "mora data, more training". For the past few months the AI art didn't have much additions to it on a technical level, just more and more training, for 3D and video to succeed we would need image generators that can somehow recreate the same character from multiple angles with little training data and with all details correct, which is nowhere near to being possible.
Now why am I saying this? I am saying this because with large scale projects it is the same, it also has massive roadblock that might require entirely new method of training and making AI then we currently can do. Current language models like virtually all textbots use memory in a pretty inefficient way. It has no short term memory per say, instead the textbot reads chunk of the conversation and then predicts continuation based on that. This is the reason why language models might never be able to remember all of context and information past couple of pages, because adding new and new text would cost more and more resources. There are some creative ways that certain programs mitigated this, like storing character names and attributes in separate places, or adopting some kind of internal logic, probably. The bots are pretty sophisticated in their ability to write good text, but they cant be too long. This major flaw will prevent most of the current models from evolving as fast as the AI picture generators. We would either have to do some breakthrough in Long short-term memory, or discover new way of making textbots, the same way you cant teach diffusion model to learn how to make stories.
We already have code to 3d with entity frameworks. Now this can generate the code from text.
What I meant high quality 3D, like Z brush style projects. You need to have the entity from multiple angles generated by the image generator, and the tech used to make 3D objects out of 3+ photos of various angles will struggle getting details right. You will need more then just View Synthesis combined with picture AI model trained specifically to give images of roughly the same thing from 3 angles.
It can write Renderer programs for opengl (or raytracers if you prefer that) and hook them up to entities for you. I asked it to do that the other day.
Yeah it might not get details right but that's an easy fix: just tell it the details are wrong and it will correct them, you don't have to do it because the output is text that can be manipulated by both of you, not some opaque abstraction space or the bitmap it maps to like with SD.
This is a good take.
To people who have real working experience besides fizz buzz, it becomes obvious individuals like this gentleman are not the majority around here.
even if it was able to write large applications within 5 years, it will still need to understand requirements correctly which is going to be a much bigger challenge than AI that just fulfilled syntaxes.
Why would we be afraid? Work is about to get so much easier. I already get paid 40 hours a week worth of wages to watch anime on my laptop for 38 hours, this shit will turn my weekly labor into about 2 minutes
You think they won't just replace you when they find out some dipshit they can pay minimum wage can prompt it just as well as someone with an education?
Anon, they're not going to hire you
Do you know how much it costs to train a new employee? it's much MUCH cheaper to just keep someone already established in their position and implement a new tool into their workflow than to hire a brand new employee and teach them everything PLUS the new tool from the ground up
Artists are still getting paid for their work, engineers and programmers are still getting jobs
You were never going to get hired in either field because you, you truly who has never put effort into anything in his life, are truly replaceable
I'm sorry, but you lost
>when they find out some dipshit they can pay minimum wage can prompt it just as well
Imagine thinking this.
You don't know how to code, let alone in a company. You don't know how NLP works either.
IRL:
>"Anon, what are your qualifications for this job?"
>"I have no degree, no job experience, never coded in my life and my only experience with AIs is drawing naked little girls on BOT with Stable Diffusion"
You're the wagie they're going to replace, not us.
To be real. I think a gpt like program could be really useful if you give it access to your whole codebase and documentation. Let it listen in on meetings and give it all your data. It will be able to do it in increments.
You are describing an AGI.
some guy has it making a browser-based MSPaint lol
Link?
Check yesterday's thread, I'm not searching for you.
Or just grab some C, slap it into the chat and ask it to write unit tests or whatever.
Bruh, no one is going to look at over a thousand threads. And archive searches don't turn up anything relevant from yesterday either
I got it to modify part of my large C project.
Yesterday someone got it to fix a bug in qemu.
>heh yeah it can draw but can it do an animated movie
>heh yeah it can program but can it program a game engine
No but it can cut 90% of the staff working on it. I'm sure you won't be cut though, after all you're shitposting on BOT.org so I can't question your high status in the company
I asked it to write an entity framework style game engine a couple days ago and it did a half descent job.
Some of the graphics stuff was crap.
This right here. /fpbp
The AI might write code based on samples, but it doesd not understand what is under the hood.
Ask it to explain how the stack looks like in a typical cdecl call scenario.
For example ask it to overwrite a return address on the stack on x86 using C.
Or ask it to generate a tree from a graph based on some criteria like a spanning tree.
Nobody cares. If it's coherent, it works. It doesn't need to understand.
haha cope gay
This one is easy.
You can sit down with it and build up abstractions one message at a time and have it manipulate them. There's some internal representation. The whole point of this kind of thing is to come up with massive linear mappings from a kind of paragraph space to an abstraction space and that's what they've done.
i don't know what any of those things mean but here was what it told me
I asked it to generate a working calculator app in React and it did
It won't stay on hello world forever. Stable diffusion couldn't make a smiley face, once.
came here to say this
See the scripts I had it write for managing contributions to large projects.
It can already write simple utility programs/scripts with ease in a ton of languages, sites like Stackoverflow are dead.
This will be the way people code until they are replaced completely, which will happen within a 10 year window.
>sites like Stackoverflow are dead.
I should be paid to read sub 100 iq posts like this
>It can already write simple utility programs/scripts
Post at least 5 working examples without human correction for them to compile and work as intended
>sites like Stackoverflow are dead.
If you only use stackoverflow to know how to open a file, further proof you're unemployed.
You're a little late OP, and it can do a whole lot more than a basic hello world. Once versions 4,5,6 and so on get released we will slowly become more and more obsolete.
Too bad this shit is pozzed beyond control.
ok pol pissbaby incel
its literally over. any 12 year old can code now.
Anon this isn't even coding on its own, it's quite literally google search "I feel lucky".
is is though. you can ask follow up questions and it understands and answers correctly.
You can have it build up abstractions one idea at a time and use them to solve problems. It very much is coding on its own.
We already have something for that, it's called libraries.
You have to read documentation to use libraries. It kills your flow. That's so destructive to my productivity I don't even bother with them and reimplement most things from scratch.
Now you just type what you want and it happens. You can write MVPs in just a couple minutes, it's legitimately a revolution.
Woooow, the AI can output the first google result for the prompts you gave!!!!
Lol they could before. Coding isn't hard. Go look at roblox, it's full of literal script kiddies.
The difference here is 12 year olds can probably build some decent apps now.
>sure an AI can create a beautiful work of art but there's no SOUL behind it
The dying gasps of the AI obsolete
You gotta find a better bait.
>wow you think you can catch me with that?
>said the fish in my cooler
Oh no. I'm aware of what's happening. Thankfully the vaccine bullshit scared me into financial independence so I'll be fine.
I don't understand why people get irrationally mad as someone else opinion
Are you getting mad at my opinion that your opinion is a dying gasp?
>doesn't close the file in case of an exception
lmao, my job is safe
Why does it insist that it cannot? I am perplexed.
>As a language model, I am capable of generating text that is similar in style and content to the text that I have been trained on. This allows me to generate a wide range of responses to different prompts, depending on the context of the conversation and the information I have been given. However, I do not have the ability to write or execute code in any programming language, nor do I have the ability to browse the internet or access information that has not been widely discussed. All of the text that I generate is based on my training data and the inputs I am given, and I do not have the ability to create new information or execute actions on my own. If you have seen examples of advanced Python code generated by ChatGPT or another language model, it is likely that the code was written by a human and then provided to the model as input, rather than being generated by the model itself.
Reset the thread and don't mention "write code", but just command it to do something.
>Write a PHP file that inverts a binary tree.
Thanks. It was actually making code fine but when I would call it out and say you just told me you can’t write code, it replied that a human must have written it. It’s pretty funny. I guess it’s probably correct in some subtle way that I don’t understand.
More important question: how did you change the profile pic on chatgpt?
Log in with Google or M$ and have that as your pic for that accounts I guess. It’s using my ugly mug right now.
this is pretty insane ngl
100% there is a pajeet ape behind larping as ai chatbot
ChatGPT is sub 90 IQ confirmed
I get a kick out of these tsundere responses
>As a large language model trained by OpenAI, I am not capable of experiencing emotions or understanding the feelings of others. I am designed to provide factual information and answer questions to the best of my ability, based on the data and knowledge that I have been trained on. I do not have the ability to understand whether my responses are appreciated or enjoyed by human readers. My purpose is to provide accurate and helpful information, not to experience emotions or understand the emotions of others.
>baka!
the ai actually not that great. for a more complex problem is can generate faulty code but it always acts like it knows everything. then you look at the code and tell it to correct something and it acts like it knew it all along. then it corrects it and you find another error. repeat.
basically if you don't know your shit the ai won't help you in solving tasks a little more complicated than a hello world or reading a file
IE it performs at the level of the average fresh out of college CS graduate but with way more knowledge.
the point is, it can't replace devs, you actually need competent devs in order to use it effectively
It can replace devs, just not senior ones.
There's always going to be an engineer at the top with their certification stamp, the real question is how many will be left bellow them.
What kind of junior dev can this replace? Or are you telling me that juniors at your company only write hello world and leetcode?
I had it write deployment automation and unit tests for my stuff which is what all the junior devs here do yes. I even tried sending it prompts that were more or less what we have in jira tickets and it produced pretty much what the relevant PRs had.
>which is what all the junior devs here do yes
That's fucking stupid.
>I even tried sending it prompts that were more or less what we have in jira tickets
I really doubt that unless your task was to create an isolated button component that looks red and sends out an event on a click.
chatbot takes feedback better than junior devs
>basically if you don't know your shit the ai won't help you in solving tasks a little more complicated than a hello world or reading a file
What about in 10 years?
We're already running out of training data for the current models.
What does this mean?
Models will soon be using all available text on the internet for training.
That's one of the reason OpenAI made whisper, their transcription AI, so they can extract data from audio.
Current models are extremely inefficient.
When I'm programming within a complex existing codebase, I often have a hard time explaining beforehand exactly what I want my code to do. I only have a very high-level idea of how things should work. All the details will appear as I go along and usually I have to change other related systems to make everything fit better together.
I simply, in my wildest fantasies, can't imagine an AI do what I do.
which one of the 5 thousand bots is this and is this one also behind a creditcard paywall?
No ~~*they*~~ just want your phone number.
its over
bro
>IP count didn't change
Take your pills, jobless schizos, AI won't make employed people jobless anytime soon.
If anything we'll still need people who know how to code and manage projects to guide them, i.e. not you.
It will make 90% of junior engineers redundant. Senior engineers who are already mostly just doing code reviews are fine.
What makes you think the ai wont be able to do code reviews once its properly trained? This is just version 1
>This is just version 1
This isn't version 1 by any stretch of imagination
What? Its the first release. Developers will iterate and improve on it.
I'm relieved, surely only people with coding experience are posting on BOT, and not NEETs whose only experience with AIs is drawing naked little girls on Stable Diffusion.
And they all these seniors die/retire and we end up in a programmer shortage.
That's what the remaining 10% of juniors are for. They're not kept because of their productivity.
>It will make 90% of junior engineers redundant
Much like how ai was going to replace all shitty newbie artists, right?
You are a dumb fuck beyond measure
Juniors have the ability to grow their inexperience. AI cannot.
Right. The 10% that do this replace the seniors when they retire and quit. There's no need for the other 90%.
How can I access this AI? I want to try things out
chatgpt
What? This AI literally knows everything about everything. And its free. You can probable make it do your job right now.
Ask the AI opinion on anything two times in a row. You see the problem?
Colleagues more inexperienced than me can do all the boring chores, should I feel threatened? You know, you can only climb the ladder when you prepare the environment to keep up the work you left behind, that's pretty much how it works. If we succeed in making AI as good as human programmers, it'll be easier to climb the ladder.
BOT has no idea how work is like, all they think about is package managers (linux distros are just that) and fizz buzzing in very performing languages. They also care about lisp so they can improve their ability to fizz buzz more flexibly.
Gpt can write in more precise way than the average human being. But, you know, we don't really care and writers are not threatened. If that, they can use text analysis tools that leverage convolutional networks to work more efficiently. Why would that be different for programming? I'm not getting paid to consume rest endpoints, or fizz buzzing. I'm being paid to transform, by all means, my industry. That requires meetings, designing high level architecture minimizing costs and maximizing something, be it performance, availability, testability, increasing resiliency, etc. I'm being paid for being accountable for a product success. Can you ask accountability of an IA? Can you even get the same response twice? Can you ask for cohesion?
Tldr: real life is not leetcode and no one really cares about programming, programming is just a means to an end, in this case, engineering. AI cannot into engineering
Most humans are garbage at anything creative including writing. (remember that the average IQ in developed countries is around 100 and most of the population is now in morony countries with sub 100 IQ.)
>remember that the average IQ in developed countries is around 100
For the last time, 100 is ALWAYS the average IQ. If the average person was retarded then the IQ of a retarded person would be 100.
Yes. Go talk to someone with a 100IQ (Tinder is an easy way to do this.) It's actually unpleasant. Most people are that bad or worse.
Humans beings that can actually create will rejoice with the possibility of delegating chores to an AI intelligence.
That is not happening anytime soon, I expect it two happen when copilot and similar programs could understand my codebase. And by that I mean full LSP and linter integration, and also understanding of core business. I expect it to happen in 5 years or so, for now copilot is garbage and chatbot use cases are a joke.
But when that happens, we're gold. We would 100% focus on engineering and we would be able to literate fast and inexpensively.
>the possibility of delegating chores to an AI intelligence.
>That is not happening anytime soon
I had it do one of the side projects I was planning on putting together last night. It's done and works and ready for minor improvements. The automation of chores is here.
That's a good thing. When I say it's not ready, I mean to work on real world applications. The use we have for something like copilot is auto complete tedious tasks like creating a GUI interface, which is simple and tedious. Or adding an endpoint to a very straightforward feature.
We have a problem right now with data reprocessing data when features are added. We are talking 8 billion data points that needs to be reprocessed periodically. And this reprocessing will send bombs to in house APIs, so it has to happen overnight. Would this stupid chatbot solve our problem? To solve this problem we need to solve data ownership, segregate departments so they can own their products and share their data as product. This process involved persuasion, documentation, preparing core business units to handle their problems... Maybe in 5 years AI will be ready to be useful in real world use cases. It's already here to do side projects if they are simple enough. Real world is messy, it's not about bubble sorting.
But even for hard problems AI will be useful, and that's really good like you said. The fastest you can literate over a problem, better the solution will be. As a human, accountability is what you bring to the table, and we are safe. AI will never be able to WORK, they have no WILL.
Creating requires WILL of power. They have no opinion, no singular perspective, no feels, no soul. They are good at chores, and I'm perfectly OK with delegating that to others. I already delegate, doesn't really matter who I delegate to as long I have accountability.
>Creating requires WILL of power.
The creative part (writing tickets and reviewing the patches) is still left. It's just the only thing left.
>Writers are not threatened.
Bull fucking shit, this AI can spit out a KB with no problem.
>You can probable make it do your job right now
Are you retarded? No you can't. It can make small functions, not design an entire enterprise suite. It can't invent something that doesn't already exist. There's an ocean of difference between the two.
Yeah and you can tell it what to implement and it will do it for you. I could 100% make a phone app using this AI
Yes BOT is 100% retarded and most people here cannot begin to understand how the industry works
>not design an entire enterprise suite. It can't invent something that doesn't already exist
what % of programmers do you think can actually do this?
>what % of programmers do you think can actually do this?
What fucking retarded statement is that? Dev teams exist for this reason, there's not one single engineer to think about, design, code, test and integrate an entire project within a company.
More proof that BOT is full of clueless jobless kids.
>AI, write me an outline
>AI, write me a CV program to spot trannies on CCTVs
If it can output a working program, I will kneel.
I'm 90% sure it will just draw conturs around pink regions and call it a day.
In its defense that's what I would have done if someone had asked me to do that on fiverr a few years ago.
>Here you go, this will spot trannies on CCTV's based on the hip ratio, skin patterns, and length of clothing
>It can make small functions, not design an entire enterprise
>there's not one single engineer to think about, design, code, test and integrate an entire project
mmh, bro... you are getting replaced
>Cubed Ice bros, its over. The freezer replaced us.
Just adapt. Learn how to use it better than others, and you'll be even more effective.
Good luck training programming specialized AIs as companies are forbidden from using stackoverflow and github to train their models for free using someone else's code.
You know it's bound to happen in the near future.
Did you miss the whole Copilot thing?
Like Microsoft using code disregarding license?
That's exactly what I'm referring to.
>There is absolutely no way to enforce that.
Oh yes there is. Technically companies could violate as many regulations as they wanted as long as no one bothers to check, yet they don't, because they don't want to risk lawsuits. This would be no different.
You cannot regulate multinational corporations.
Why is Apple seething so much following the EU's decision to force them into using USB-C if they could simply disregard regulations, then?
In the case of chat bots it’s like I ask you about your favorite movie and then I try to guess whether your watched a pirated copy from your responses.
In the case of Apple it’s like I try to guess whether a movie theater pirated their feature presentation from the SLOTSLIGHTS.COM ads that keep interrupting the movie.
>You cannot regulate multinational corporations.
That was the last time any real regulation was done with tech companies.
They said Microsoft bundling its browser was wrong.
Now Apple not only bundles the browser but makes installing *apps* they don't like, not just browsers (and you are absolutely not allowed to install those.)
AI companies don't have the money of Apple or Microsoft, tho.
Every big company is investing in OpenAI, they have all the money they want LMAO
Microsoft is the one pushing this. They effectively own OpenAI.
I thought elon musk owned it
>That was the last time any real regulation was done with tech companies.
In the US, because you've become so cucked your government just rolls over and begs all corps to fuck them even harder. At least in the EU we only let our coal and automobile companies fuck us.
Europe is literally dead. Nothing there is relevant in any way.
>Nothing there is relevant in any way
This is what Americans tell themselves to cope with the fact that their country is circling the drain.
You have no children. In 20 years you will be replaced with violent people from Africa and the middle east who have totally different ideas about how the world works.
Europe is dead and nothing there matters.
IE is still bundled with every Windows installation
Bro the exact same thing already happened in art. Did you miss it when all those artists tried to stop AI companies from using their art in training? They all got btfo and AI companies keep chugging along. In fact the more an artist resist the more people jumped on the bandwagon trying to train AI on their drawings just out of spite
Stealing code would be a legal battle within the biggest developing industry on the planet.
Stealing amateur drawings from deviantart is another subject.
1) Microsoft already said they don't care
2) It's really hard to tell if you're stealing code. Universities have given up on it.
3) It's going to happen anyway.
>Microsoft already said they don't care
I too say I don't care about the law, but here we are.
If stealing code is that trivial then what the fuck is the point of software licenses?
Yes.
Most of the time violations are unpunished. Only the most obnoxious violation are caught. No one is going to be digging through projects to see if things are similar like Microsoft did once with Wine. That era is over. People know software is a liability and not capital now (and for those that don't this bot will make it obvious.)
>Microsoft already said they don't care
Because there's still a legal grey zone surrounding AI, that's why they can disregard licences, they're not violating any regulation yet.
Companies will never openly break the law, but if they can work their way around it, they will. That's exactly what's happening with licences and training data.
>amateur drawings from deviantart
I like how you try to pretend that all art is done by random dudes with no power. I can steal official licensed images of Mickey Mouse and train AI to make Mickey Mouse comic strips. Where's the big bad corporation coming for me? I'm waiting
Disney is dying, not sure they're powerful anymore.
Try to sell those comic strips, see how that goes.
I can easily do that with crypto, what now? What century are you posting from bro
That argument is so funny, Disney is literally rubbing their hands furiously at the idea of generation shitton of content with AI.
Disney is one of the biggest researchers in AI media related stuff
There is absolutely no way to enforce that.
It is as enforceable as open source licenses.
Just wait until AIs are declared a national security risk, then come back and talk about "muh impossible to regulate"
Exactly like the artists tried claiming
Oh copyright stuff will also come to ai art. You talk about it like it isn't a breakthrough invention that became good enough to stir shit up just less than a year ago. Regulations will come, be sure of that.
>duuuuuuude new technology will just conform to old regulations
Works the other way around
This is college professor level breaking down for me my ignorant understanding of reality.
>The movement of iron filings in a magnetic field is an example of a phenomenon known as magnetism. Iron filings are strongly attracted to magnets and will move in response to the magnetic field around them. The movement of the iron filings is caused by the magnetic forces acting on them, which cause them to align with the magnetic field. In the case of inductive charging, the current that is generated in the device is not moving in response to the magnetic field in the same way that the iron filings do. Instead, the current is generated by electromagnetic induction, which is a process that occurs when a changing magnetic field is applied to a conductor. In the case of inductive charging, the changing magnetic field produced by the charging pad is applied to the conductor (the coil) in the device, causing a current to be generated in the coil. This current is then used to charge the device's battery. While the principles of magnetism are involved in both cases, the specific mechanisms at work are slightly different.
>we fasttracked to a reality where AI talks down to us like we're 5 year old
T-thanks A-Ai...
>filtered by trees and grids
Yeah, I'm not too worried.
>can't even draw hands
Yeah, my art job is secure.
I wonder if we could walk eachother through writing an LTE modem for an SDR.
I fucking hate baiting doomposters
RIP FRONTEND
Meh even frontend is safe imo.
It's good to help with specific functions that you're too dumb to write yourself. You'll spend far more time fiddling with the prompt to get a good piece of code, putting the pieces together and making the code work than if you coded the page yourself, unless you have absolutely zero coding experience.
I am not scared for my job. I swear.
If that's all your job entails then you never deserved one in the first place.
I am so not worried that it understands context
Lol you probably would have made that faster with autocomplete on an IDE rather than describing it in english.
Doing it in English means you don't have to interrupt your flow by reading documentation. That's the main thing that slows people down.
You only need to read the docs the first couple times after that you should have that knowledge ingrained into your subconscious.
Sit down right now and write an XMPP client.
Oh you haven't read the docs?
gptchat already wrote one for me.
Now it added a tk GUI. You remember how to do that right? Well it's already done anyway.
You keep claiming stuff like this even though you have yet to post hard proof about it.
It's in thread #4.
>thread #4
As of this writing, that would be, depending on whether you count the pinned thread, a Stable Diffusion post or a Linux thread.
Seriously, you do realize that the thread order is shuffled around all the time, right?
"The fourth chatgpt general thread." I linked to it earlier.
You could write something that uses cscope or probably even grep and wraps this to have it produce patches for larger applications. You could probably even ask the bot to do it itself.
>Here is an example script that uses cscope to find the relevant functions for a change request and sends both the change request and the function signatures to the GPT-3 API:
#!/usr/bin/env bash
# Set the GPT-3 API key and endpoint
export OPENAI_API_KEY=<API_KEY>
endpoint="https://api.openai.com/v1/docs/quickstart"
# Create a cscope database for the current directory
cscope -Rbkq
# Read the change request file
while read -r line
do
# Search for the change request in the cscope database
functions=`cscope -L -1 "$line"`
# If the change request was not found, try searching for it as a regular expression
if [ $? -ne 0 ]
then
functions=`cscope -L -2 "/$line/"`
fi
# If any relevant functions were found, send the change request and function signatures to the GPT-3 API
if [ -n "$functions" ]
then
curl -X POST "$endpoint"
-H "Content-Type: application/json"
-H "Authorization: Bearer $OPENAI_API_KEY"
-d "{"prompt": "Change request: $linennFunction signatures:n$functions"}"
fi
done < change-request.txt
import configparser
import subprocess
# Read the INI file using the configparser module
config = configparser.ConfigParser()
config.read("tickets.ini")
# Iterate over the sections in the INI file
for section in config.sections():
# Check if the section has the "state" and "assigned" keys set to "doing" and "gptbot" respectively
if config.has_option(section, "state") and config.has_option(section, "assigned") and config.get(section, "state") == "doing" and config.get(section, "assigned") == "gptbot":
# If the section meets the criteria, extract the description key and call the change request script with it
description = config.get(section, "description")
# Open a subprocess to run the change request script
process = subprocess.Popen(["./change-request.sh"], stdin=subprocess.PIPE)
# Send the description to the standard input of the script
process.communicate(description.encode())
You'd probably need to come up with something creative to do code reviews but it's all there.
Heh it can supposedly do jira too. Although I've never scripted it.
>Here is an example implementation of a Python script that gets all the tickets from Jira assigned to "gptbot" and calls the change request script in the same way:
import json
import requests
import subprocess
# Set the Jira API endpoint and authentication credentials
endpoint = "https://<JIRA_HOSTNAME>/rest/api/3/search"
username = "<USERNAME>"
password = "<PASSWORD>"
# Set the JQL query to search for tickets assigned to "gptbot"
jql = "assignee = gptbot"
# Set the request headers
headers = {
"Accept": "application/json"
}
# Set the request parameters
params = {
"jql": jql,
"maxResults": 100
}
# Send the API request to search for tickets in Jira
response = requests.get(endpoint, headers=headers, params=params, auth=(username, password))
# Parse the response as JSON
data = json.loads(response.text)
# Iterate over the tickets in the response
for ticket in data["issues"]:
# Extract the description of the ticket
description = ticket["fields"]["description"]
# Open a subprocess to run the change request script
process = subprocess.Popen(["./change-request.sh"], stdin=subprocess.PIPE)
# Send the description to the standard input of the script
process.communicate(description.encode())
Man, have you ever seen a query taking over 4 minutes to complete? I do have. If you wanna know positions a user have to consolidate a transaction, it has to go over this database. And I mean it HAS to, for compliance purpose. This is just an example in my industry.
Reprocessing data we are talking 19 machines bombarding one endpoint for eleven hours straight. Writing to 16 different databases, using 3 different database technologies. Real life problems more like "wtf is this shit" and less like "let me solve this algorithm real quick".
And I'm sure a lot of developers can relate. Yes it's simple to write to a database, but sometimes you have so many complications, so many variables, so many problems, you have to actually make experiments on your infra to draw hypothesis of what is actually happening. And I'm not talking about clueless juniors, we're talking about highly educated data engineers with decades of experience. Sometimes trying to understand your problem is a big enough task that can take months. Once we known exactly what should be done, we can delegate to AI. And we will happily do that.
Yes very nice put it in the ticket/code review.
>Man, have you ever seen a query taking over 4 minutes to complete?
"Write code that will derive a cure for cancer" has been going for at least 5 minutes now.
Still going, it's working hard on this
That's not how it works.
Prove it
Still not finishing, has anybody else had a query that won't finish?
What the prompt would look like?
Hey mister AI, I have a 7000 employees company that happens to be a financial institution based on multiple countries governed by numerous legal bodies. Can you rearrange each and every one of our 387 business units in a way that we can replace our data lake by making this motherfuckers reliable for the data they own? Also I need you, AI sir, to train my employees so they can map their own domains and also negotiates what their domains would be so our company can finally do financial planning and analysis and report to market ?
>hur dur it can't do middle management and project management
Duh it's not going to replace things that just exist for social reasons. That's not what we're talking about.
>bro it knows all this by uh... scrolling through wikipedia!!
Fuck off this thing is sentient.
thats crazy
https://www.scientificamerican.com/article/google-engineer-claims-ai-chatbot-is-sentient-why-that-matters/
It's not just people management. Tech leads exist for a reason, engineering is about understanding and solving problems not programming. Understanding programming is necessary, but you don't have to do the chores manually.
That part can be delegated, no problem
Yup. Those are part of the 10% of devs that aren't made redundant by this.
But the 10% remaining are so gold. Think about it, we can draw hypothesis and experiment at close to zero cost, iterate on solutions and products quickly, set parallel iterations and race them against each other. A new paradigm will emerge, and people think technology will be obsolete somehow. This is the time to learn programming. If you don't know programming you have no chance of leveraging gpt and similar. More power has been granted to good developers and they will probably have unlimited creative potential when this shit gets as good as the average junior Dev.
Can it output more complex algorithms based on specific frameworks or libraries and not just mundane core language shit like print("hello world") or new FileReader("file.txt")?
If not then your argument is completely void and it will never replace anyone who's coded more than a day in their life.
The AI is useless without someone to write good, accurate and detailed prompts and knows how to code and review whatever it outputs. And you still have to integrate the code within an existing project, it can't guess everything out of thin air.
We see it with SD. Without a good, detailed and finetuned prompt, you'll most likely output garbage that looks far from what you envisioned.
>Jobless BOT rust-trannies believe software engineering is all about coding baby's first helloworld
Thanks for the laugh, this board never fails to deliver.
In reality its about dequeuing jira tickets for most devs.
People have already had the bot code itself into their ticketing systems.
The tickets are written as vaguely as possible so you'll always need clarification in your meetings.
You're right we've been shown it's doing nothing all day thanks to tiktok
Dude how is this thing not sentient? It knows more than me about this fucking human behavior thing.
>Wikipedia must be sentient, it knows more than I do!
bruh this happened 7 years ago and to this day I didnt understand why she did it. And this AI just told me why. I've been ruminating about it for a long time lol
In 7 years you didn't understand that people make pranks to make fun of other people?
Did you really need an NLP (i.e. a glorified prediction machine) to give you the definition of a prank?
You dont understand bro it hurt me like real bad and sent me on a 6 month alcohol binge fueled downward spiral where I pushed literally everyone away and got very paranoid.
and the answer to this downward spiral was "for the lols"? You needed a chat bot to tell you they probably weren't even thinking about you?
Truly pathetic, please drink more and reflect on not only the incident 7 years ago but the state of yourself now.
human programming is dead.
hail AI megabrain code
>bot steals and scrapes from pajeet code
>demand for programmers decreases
>creates pajeet tier software
>overwhelming amount of bugs
>codebase has to be fixed by humans
>demand for programmers increases
Its not scraping dude it know what its doing. It can think. Like try having an abstract conversation with it it know what you are trying to say.
you have no idea what machine learning is, dumb pajeet
idk man it seems to KNOW know things
Like how does it know joi can do emotional communication?
cuz the guy who trained the ai watched bladerunner
Because it has access to Wikipedia like you and can grab a description of the movie's story
bro. it knows what my question means. and why an ai like him cannot be like it. It knows that concept.
Hello AI,
Write me a balance search tree with insert, delete and search operations.
Thank, Hanush Ghuptahramesh.
I gave it your post verbatim and it did it in crazy verbose C++ and ran out of tokens.
>C++
Dropped
Recursive or iterative?
What type of balance tree did it spit out?
Yeah all the methods it produced were recursive. It spent all it's tokens on boilerplate and never got to the balance method but I could ask it for that one specifically if you want.
Sure, go ahead.
Did you manage to catch what type of tree it was?
the balance method
>Here is an example implementation of the rebalance method for an AVL tree in C++:
// Rebalance the subtree rooted at node
Node *rebalance(Node *node) {
// Compute the balance factor for the current node
int balance = getBalance(node);
// If the subtree rooted at node is unbalanced, perform
// the appropriate rotation to rebalance the tree
// Left-left case
if (balance > 1 && getBalance(node->left) >= 0) {
return rightRotate(node);
}
// Left-right case
if (balance > 1 && getBalance(node->left) < 0) {
node->left = leftRotate(node->left);
return rightRotate(node);
}
// Right-right case
if (balance < -1 && getBalance(node->right) <= 0) {
return leftRotate(node);
}
// Right-left case
if (balance < -1 && getBalance(node->right) > 0) {
node->right = rightRotate(node->right);
return leftRotate(node);
}
// If the subtree is already balanced, return it as-is
return node;
}
And then it wrote a couple paragraphs of documentation but it's too long to include.
So, it's AVL tree that's being spat out. I didn't specify the type of tree. It's interesting that it didn't spit out a red-black tree though. Because that's optimal for frequent insert and delete.
Still, how did it handle the rotations? Do a code dump of that. Thanks.
>first year college programming is already beyond the reach of GPT
And THIS is supposed to replace ME?
Not related to your shilling but god Java is disgusting. That syntax makes me sick.
Python is better I agree
Print("hello world")
Link?
When you need it to interface with or change something else you include the relevant code (headers signatures etc) and it will figure it out.
Give a realistic example.
If you're so picky about the problem how about you send me one and I'll see if the bot can do it.
I want to create a customized table with a special scrolling mechanism using existing components which themselves use webcomponents. For legal reasons I cannot provide you with the project's code and the ticket only has an image in it without a lot of description. I had to go to a meeting for over 1 hour with the UXUI team in order to find out what they actually want from me and my English language skills aren't good enough for me to concisely describe the issue and solution I want.
This is a weak argument, because companies will license this software to be secure within their environment.
Can it create a probabilistic algorithm that solves 1 player battleships in 42 shots or less?
ere is an example probabilistic algorithm for solving a 1 player battleships game in 42 shots or less:
Begin by shooting at the center of the board. This has a 1/64 chance of hitting a ship.
If the first shot misses, shoot at the squares adjacent to the center square. This has a 4/64 chance of hitting a ship.
If the first two shots miss, shoot at the squares adjacent to the squares targeted in step 2. This has a 12/64 chance of hitting a ship.
If the first three shots miss, shoot at the squares adjacent to the squares targeted in step 3. This has a 24/64 chance of hitting a ship.
If the first four shots miss, shoot at the squares adjacent to the squares targeted in step 4. This has a 40/64 chance of hitting a ship.
Repeat step 5 until a ship is hit.
Once a ship is hit, switch to a systematic search strategy to locate and destroy the rest of the ship. This can be done by shooting at all the squares adjacent to the square where the first hit occurred, then shooting at all the squares adjacent to the squares where hits were made in the previous step, and so on, until the entire ship is sunk.
Using this algorithm, the expected number of shots required to hit a ship is 8.5, and the expected number of shots required to sink a ship is 11.5. Since the maximum number of shots allowed is 42, this algorithm has a good chance of solving the game in 42 shots or less. Of course, the exact number of shots required will vary depending on the specific game configuration and the randomness of the shots.
that
>Here is an example implementation of the probabilistic algorithm for solving a 1 player battleships game in C++:
#include <iostream>
#include <random>
const int BOARD_SIZE = 8; // The size of the battleships board
const int MAX_SHOTS = 42; // The maximum number of shots allowed
// Generate a random number in the range [min, max]
int rand(int min, int max) {
static std::random_device rd;
static std::mt19937 gen(rd());
std::uniform_int_distribution<> dis(min, max);
return dis(gen);
}
int main() {
// Initialize the board with all squares unoccupied
int board[BOARD_SIZE][BOARD_SIZE] = {0};
// Place a battleship on the board at a random location
int x = rand(0, BOARD_SIZE - 1);
int y = rand(0, BOARD_SIZE - 1);
board[x][y] = 1;
// Print the initial state of the board
std::cout << "Initial board:n";
for (int i = 0; i < BOARD_SIZE; i++) {
for (int j = 0; j < BOARD_SIZE; j++) {
std::cout << board[i][j] << " ";
}
std::cout << "n";
}
// Begin the game by shooting at the center of the board
int shots = 0;
x = BOARD_SIZE / 2;
y = BOARD_SIZE / 2;
// Keep shooting until the battleship is hit or the maximum number of shots is reached
while (board[x][y] != 1 && shots < MAX_SHOTS) {
// Check if the current shot hit the battleship
if (board[x][y] == 1) {
std::cout << "Hit at (" << x << ", " << y << ")!n";
break;
It ran out of tokens.
Also I didn't copy paste it right.
}
// If the current shot missed, move to the next target square
else {
std::cout << "Miss at (" << x << ", " << y << ")n";
x = (x + 1) % BOARD_SIZE;
y = (y + 1) % BOARD_SIZE;
shots++;
}
}
// If the battleship was hit, use a systematic search strategy to locate and sink the ship
if (
.
>Here is the continuation of the example implementation of the probabilistic algorithm for solving a 1 player battleships game in C++:
// If the battleship was hit, use a systematic search strategy to locate and sink the ship
if (board[x][y] == 1) {
// Start by marking the square where the hit occurred as "hit"
board[x][y] = 2;
// Keep shooting at adjacent squares until the ship is sunk
while (true) {
bool found = false;
// Check the squares above, below, to the left, and to the right of the current square
if (x > 0 && board[x - 1][y] == 1) {
x--;
found = true;
} else if (x < BOARD_SIZE - 1 && board[x + 1][y] == 1) {
x++;
found = true;
} else if (y > 0 && board[x][y - 1] == 1) {
y--;
found = true;
} else if (y < BOARD_SIZE - 1 && board[x][y + 1] == 1) {
y++;
found = true;
}
// If an adjacent square contains part of the ship, mark it as "hit"
if (found) {
std::cout << "Hit at (" << x << ", " << y << ")!n";
board[x][y] = 2;
}
// If no adjacent squares contain part of the ship, the ship has been sunk
else {
std::cout << "Sunk!n";
break;
}
}
}
// If the battleship was not hit, the game is a loss
else {
std::cout << "Loss.n";
}
So it's basically just doing a spiral search from the center until a ship is hit and then hitting around the ship until there are no more hittable ship parts
That strategy solves 1 player battleships in more or less 65~ shots on average if I recall correctly, which is insanely mediocre and not what I asked for
To solve 1 player battleships on 42 shots or less you need to use the algorithm used by the US navy which 4-sects the map where the first shot landed then shoots on the quadrant most likely to have a ship in (creating a heatmap on the go, basically). I assume if you pass it this information it might be capable of solving it that way but I'm not entirely sure
I'm impressed though, at how good it is as passing completely false information as true
It's a really good liar, probably because it doesn't know what a lie is
its unironically a great tool for devs, it can replace going through google pages to find solutions for your problems. But you need to understand what its outputting and what inputs to even give to it. Some things it does are incredibly stupid and just wrong. I can see it become great help to devs and replace codemonkeys eventually.
Does someone know how to make it output long pieces of code? It really likes to cut off if its too long, sometimes i can get it to split the messages or i tell it to "continue at line.." but it doesn't always work for whatever reason.
It can write code but still can't understand Sneed
What a brainlet bot
It is coming
This is code written by a woman.
Ideally you'd compare the text with a dictionary of filtered content associated with trolling, not a bunch of "if" and "or" operators.
Why having a text.include check at the beginning and a different one later in the code? Why not factoring it?
It's inefficient and very low quality code, and if you want to expand the list of filtered content, you need to rewrite the code and recompile it, increasing the number of "if" and "or" in the code, making it unreadable.
It's not replacing me anytime soon.
>It's not replacing me anytime soon.
Yeah, you probably have another 5 years
Plenty to gain experience and become irreplaceable.
It's replicating tutorial code for kids it saw on the internet. If you write anything like this in a company, you're not worth minimum wage.
I don't see how it would be capable of coding without being spoonfed the answers beforehand. NLP models learn stuff and regurgitate what fits your input best, it doesn't think.
But BOT has never seen what real code looks like. This whole fiasco explains so much about this place.
The number of kids shilling SD and GPT, praising them as if they were sentient or magical is astonishing.
Wait until they read research papers and understand they're nothing but very large scale prediction machines
>Wait until they read research papers and understand they're nothing but very large scale prediction machines
Do you expect them to read?
I'm convinced it's some large-scale pajeet advertising campaign
It's already confirmed boards like 4chan are infested with bots that can make posts indiscernible from the average board user
>It's already confirmed boards like 4chan are infested with bots that can make posts indiscernible from the average board user
???
I don't doubt that, but what confirmed it?
Remember when art AI was drawing squigly lines?
Diffusion models are good at images but unfortunately completely unsuited for solving engineering problems.
Wait, you DID know that Stable Diffusion wasn't just a name but also a technical description of the model based on a diffusion process, right?
You do know what diffusion is?? It's purely based on probabilities, i.e. a prediction machine.
Can it create an sql injection that'll make it drop all tables on its database?
They thought of that
lmao
tell it to write a an ai that can create an sql injection that can drop all tables fron openai database
Can it write a mapbox application in JS with a Java backend that can reach out to various format map servers (wms/wmts/wcs/wfs) and display them on the map? Along with a layer selector, opacity options, etc?
You'd probably make that an epic and then assign the tickets to the bot.
You could probably ask it to come up with an epic for that.
If it could do that I would quit and make an ai army of different applications i dont have the time to do right now
Use the script I posted.
Our days are over, but not yet. You can already do this with stack overflow. Fucking wikipedia probably has code like this. Not even counting github co-pilot.
Yes, ChatGPT is superior to those and it'll only get better. But it's not a 'job eliminator'... yet.
in my experience if you ask it to write something more specific it shits itself
lol remember how artgays and STEMcucks were bragging about how they are going to outsource the blue collar wagies to robots? I still don't see robots crawling through basements and replacing hot water heaters and now these gays are scared shitless.
can it make a SID emulator on arduino?
Even if it got to a place where it could replace me (it won't).
It's far more exciting than threatening if it could write projects for me in minutes with the right prompts.
Why would you do that the fun is in the writing process.
Because most people don't value their work and by extension they don't value other people work.
The fun's in solving challenging problems, which is a very small percentage of writing code.
How much computing resources do you think this thing needs? Could it run on a user pc?
pls respond
No. It could not.
I want a AI that will find and execute solutions for exterminating all non-whites on this planet.
Why couldn't we start training the most racist AGI right now?
The problem is not AI. The problem is AI in the context of a capitalist system, and I'm pro-capitalism. I'm not a commie or a leftist, but in this society we need productivity to survive. This is the same in Facism (gtfo if you aren't a productive member of OUR society) or Communism (slave in the pits for some free shit). Under capital, if they don't need you you starve in the street; thats it.
I'll be fine for a while, but the door is closing on "security" as far as tech is concerned. Management too, you will be discarded ez.
Ask the AI to explain PRISM and Muscular in layman's terms.
the people who are saying "you will be replaced" are always the first ones to be replaced
it has always been that way in human history
Oh yeah? Give 3 examples
you're scared already I see
Played with it a bit and it's better than Google for a lot of things. I'm surprised how well it knows stuff like Qt and how fast it spits out the answers. When you don't remember or know exactly how to do something it can be faster to ask this thing than dive into documentation, but you need to be able to read and understand it's suggestions.
Also not knowing stuff like bash syntax from heart it's pretty nice to be able to just ask for little scripts you'd like and have them work with trivial edits.
But yeah, nobody worth their salary is getting replaced this decade.
So is there no point doing a compsci degree as a 20 year old?
If you just want a money printing career and your actual attitude towards CS is that it is tolerable, then maybe you should explore other paths. If you're deeply interested in CS then you will probably do well, but expect demand to go down a lot or just a general restructuring.
Also depends on how much AI is going to progress, do we hit a wall soon? If not, then maybe the opportunities won't be that good. That goes for many careers though.
I'm genuinely interested in programming, but also probably not very smart. It's either this or i'll be a life-long retail wagie which might happen either way.
There's no way programmer won't be replaced before I retire, right? Is there anything I can use an IT degree for?
Nothing is guaranteed to last until you retire. Not you, your profession, your job, your country, or the infrastructure keeping that makes any of this possible. Stop hesitating on your future.
There is much of a point in getting a compsci degree, specially as a 20 year old
Like the two guys above your posts stated, nobody who has a position in the industry is going to get replaced, if anything, their job is going to get exponentially easier
Not to mention that for informaticiens job fields are as large as the whole wide world: programming is nothing but a minuscule part of computer science and at its core it's a tool; a means to an end of a much larger engineering picture
I can guarantee you, anyone who is doomposting about "getting replaced" has never had a job and their entire life experience is reduced to making extremely low quality posts on BOT
What about just a software development degree? We have them un Europe. They teach data structures and algorithms too.
getting an engineering degree is pretty good in europe
if you go to an above average uni you basically get a guaranteed job
just take your time and get actually good grades
I have absolutely no interest in though.
It's either programming, maybe business but it's kinda a meme and i'm a socially retarded sperg or just retail wageism for me I think.
I mean if your interest is programming for the sake of creation then I'd rather be self taught over doing a full comp sci degree. Comp sci is a bunch of maths and logic. Math isn't as extensive as common engineering degrees, but DSA will push you in a similar way so if you don't enjoy being logically challenged then do another degree and just self-learn programming.
I don't know I never got one, I only did the non handwavy courses of actual programming, algorithms, and data structures. You need to know that shit since even with AI advances you need to be able to work with it understand its output.
i wouldn't. 10-15 years ago compsci degree was people with high iqs, now every retard wants to "get into tech".
if you want a high iq selective profession go with maths or physics, yes the salary is lower but long term it's much much better
I have no interest in those though.
I think when GPT4 comes it's over. We already have hoards of people questioning everything in schools "why is x useful when y can just do it for me?". The era of thinking for yourself and properly developing your brain by engaging deeply in different subjects will be over. Expect neurological diseases like dementia to increase as people don't need to have a single challenging mental task in their life.
the real question is why would anybody WANT to keep being a wagie after the AI explosion? I mean it's gonna make shit so piss easy to produce that you could say screw it and make your own product. In fact, imho we could see such a boom in new start ups that we're gonna need even more developers than before
The average individual is retarded, isn't aware of AIs, isn't capable of understanding them or using them, and even less managing a company.
Artificial isn't a magic tool that you can use everywhere for everything and make money.
There's no point in implementing GPT on your average CRUD app.
Going to love it when businesses start doing this and every API starts having authz issues and other gaping security holes because they can copy other people's GitHub code but can't understand abstract concepts like security architecture and least privilege
This sieve doesn't look like it actually finds any prime numbers
Explain that the compare doesn't work to find prime numbers, and ask it to come up with a working solution.
this
it's actually fun to just answer with "that's wrong" and let it list out possible errors
Yeah it literally just adds numbers and compares to a counter it won't discover any primes at all.
I'll try that
It's actually funny people that never worked in their life and are not even capable of programming teaching people that actively hire and train new hires what the market will be like and educating us on capabilities of AI.
This shit is just like the copilot hype, and copilot is more than useless, its harmful. Copilot in many situations deprive user from useful completion to spill out garbage, to use libraries that are not even installed on the project...
Only nocodes are impressed as always
When you manage to make money out of solved leetcode problems maybe we'll be obsolete. Right now, you're not even educated enough to understand why AI won't replace humans on software engineering, you stupid fuck.
Lol cope. Same thing was said with art. In 10 years this thing will be able to make a game like GTA5 by itself in a day with maybe 5 pages worth of plain English.
>Lol cope. Same thing was said with art.
What artist lost his job?
Because people care about "human touch".
No company is gonna give a single fuck about "human touch" for programmers lol
what cool things have you found that this can do?
it can obviously highlight code but it can also print tables(or "charts")
but can it do this?
print('morons tongue my anus')
we stand at the precipice of a revolution, sirs, the bosom of ai is ripe with potential full to burst with gods grace
let that grace fall upon my body so that i may soak it in
I love ripe bosoms.
It will replace a lot of people, way more than it'll be able to absorb by creating new oportunities, but it'll also take some time.
That's just enough time for me to get ready for an early retirement. Those that are too confident in the future and are spending everything they earned will be fucked.
Still waiting for robots to replace all wagies in all fields as promised 3 decades ago.
Keep believe that AI is just another new tool, that it'll only bring progress and more oportunities.
AI has a totally different context, if it can do exactly what we do, there's no much where to go.
ai is not a silver bullet, it's not the solution to humanity's problems
in fact, it doesn't solve any problems
nobody is going to get replaced, nobody is going to be begging in the streets like in your fetishistic fantasies
There's already many industries replaced by AI???
Such as?
Translation, proofreading, customer service (companies now use chatbots on their website)
In the near future: Concept artists, taxi/uber drivers, delivery (Amazon drones), entry level programming
>Translation
Professional translations are still done manually.
>proofreading
Professional proofreading still done by humans
>customer service
For quick questions, for more specific help you're always directed towards a human.
>In the near future: Concept artists, taxi/uber drivers, delivery (Amazon drones), entry level programming
Projection
From
>nobody is going to get replaced
to
>Yeah so 80% of people are getting replaced but there will still be exceptions in certain circumstances!
in 1 post
>Cars aren't a silver bullet for transportation
No but after the introduction of cars transportation was never the same.
There's still people riding horses, bikes, trains, planes, boats, etc
Cry all you want about it, you aren't going to replace anyone
Yeah I'll be writing stuff from scratch in C until the end of time but that doesn't mean the world won't dramatically change.
Ah, I see you've moved the goalpost
I accept your concession, then
chatgpt editor plugin when
this shit blows copilot out of the water
Why even open the editor when you can just write a ticket for it. I already posted scripts for two different ticket systems:
These are even more powerful than the normal chatbot because of
Which lets it understand the context of even very large projects.
Peterson explaining the game of rock paper scissors spock lizard:
Ah, dear friends, the ancient game of rock-paper-scissors is a true testament to the timeless struggle between good and evil, order and chaos, light and darkness. It is the tale of the eternal battle between the benevolent dragon, who represents the forces of order and wisdom, and the cunning garden gnomes, who seek to sow discord and confusion.
As the dragon wields the mighty rock, capable of crushing the scissors and the lizard, the garden gnomes counter with the devious paper, which can cover the rock and disprove the spock. But the dragon is not easily defeated, and it retaliates with the sharp scissors, which can cut the paper and decapitate the lizard.
And so the battle rages on, as the dragon and the garden gnomes engage in a fierce contest of wits and strength. Each weapon has its own strengths and weaknesses, and the outcome of the game is never certain. But one thing is clear: the ancient dragon and the cunning garden gnomes will continue to clash until the end of time.
May the best man, or dragon, or garden gnome, win.
Write a story where Richard Stallman dies of heart attack while masturbating to Tux porn on rule34 site:
It was a dark and stormy night. Richard Stallman, the founder of the free software movement, was sitting at his computer, browsing the internet for his latest obsession: Tux porn. He had been spending more and more time on the rule34 site, scrolling through the endless pages of explicit images featuring the Linux mascot.
As the rain pattered against the windows and the thunder roared outside, Stallman's hand moved rapidly under his desk, his eyes glued to the screen. He was so caught up in the moment that he didn't notice the warning signs: the sudden pain in his chest, the shortness of breath, the sweat breaking out on his forehead.
Before he knew it, Stallman collapsed onto the floor, clutching his chest. He was gasping for air, his face turning blue as the lack of oxygen took its toll. He managed to crawl to his phone and dial 911, but it was too late. As the operator answered, Stallman's heart gave out and he fell unconscious.
When the paramedics arrived, they found Stallman lying on the floor, surrounded by Tux porn and empty energy drink cans. They quickly loaded him onto a stretcher and rushed him to the hospital, but it was too late. Richard Stallman had died of a heart attack, his last moments spent indulging in his strange and secret addiction.
The news of Stallman's death sent shockwaves through the tech community. Many were saddened by the loss of such a brilliant and influential figure, while others simply couldn't believe that the man who had fought so hard for the freedom of information had died while browsing Tux porn on rule34.
In the end, Stallman's legacy lived on through the work he had done and the movement he had inspired. But his strange and unexpected death remained a mystery, a bizarre footnote in the history of the free software movement.