rip programming job
rip programming job
Falling into your wing while paragliding is called 'gift wrapping' and turns you into a dirt torpedo pic.twitter.com/oQFKsVISkI
— Mental Videos (@MentalVids) March 15, 2023
rip programming job
Falling into your wing while paragliding is called 'gift wrapping' and turns you into a dirt torpedo pic.twitter.com/oQFKsVISkI
— Mental Videos (@MentalVids) March 15, 2023
Yes free me
AI art didn't put human artists out of work, AI text-to-speech didn't put voice actors out of work, I think it's rational to say AI language models won't put programmers out of work either.
>AI text-to-speech didn't put voice actors out of work
I would say that SOTA voice generators rivals random throwaway NPC interjections and babble. The ability to generate lots of different voices is a plus for such work where you need tons of variation for cheap.
>AI art didn't put human artists out of work
yet
>AI text-to-speech didn't put voice actors out of work
yet
2 more weeks
>AI art didn't put human artists out of work
It's been like 2 months
>it's only been two months
Stable diffusion came out six months ago. It only took two months after ChatGPT dropped for mass codemonkey layoffs to start happening.
Are the mass layoffs the result of ChatGPT or the result of Elon demonstrating that companies can safely get rid of dead weight to save money?
You can add hiked interests rates into the mix. Money is no longer free. Also some have accused FAGMAN of simply locking up talent in useless shit to prevent competition.
https://finance.yahoo.com/news/google-over-hired-talent-fake-114331193.html
The latter is an orchestrated excuse for the former.
> Art
Human expression needed, being machine made literally devalues it to complete worthlessness
> Speech
Human expression needed or at least preferred. People would rather have voice actors than speech software.
> Code
Literal machine language, human input doesn't makes it more worthwhile or gives it any more value.
I wish humans were more valued in their crafts, but coding is much easier to replace for machines than arts or acting because of the lack of human craft involved.
In the end I would have no one losing jobs to AI, I don't care about making the tech industry even richer.
Computers aren't human and cannot solve human problems without human intervention. Such that a human would need to tell the computer what to do. Even if the AI problem does occur, it will only optimize out the unneeded.
I really hope that AIs content generators(text, images, audio) are banned.
But I know they will probably not because 'hurr productivity'. in the long run, for people's individual development, this technology is a tragedy.
How exactly do you propose banning them? Outlaw PC's and have some sort of dumb terminals that connect to supervised cloud services as the only legal way of owning a computer?
>years ago
>compilers are invented
>oh noes my porgramming jerb!
While the invention of compilers did automate certain aspects of programming, the introduction of artificial intelligence is a whole different ballgame. Unlike compilers which are designed to translate source code into machine code, AI is capable of learning and adapting to new situations. This means that AI has the potential to automate not only low-level programming tasks, but also high-level decision making and problem-solving tasks, which were previously thought to require human intelligence. As a result, AI has the potential to replace a much wider range of programming jobs than compilers ever could. So, don't be too quick to dismiss the impact of AI on the programming industry. It's not just about "oh noes my programming jerb!" anymore bro, you're comparing apples to oranges here. Compilers just translate code from one language to another, while AI can actually learn and adapt to new tasks. Plus, AI is still in its early stages and can't replace all programming jobs just yet. Don't be a doomer, keep learning and stay ahead of the game.
My point us with compilers there are more programmers now than ever. AI will change what we do, but will result in even more of us.
stupid phone keyboard. soz
>This means that AI has the potential to automate not only low-level programming tasks, but also high-level decision making and problem-solving tasks
It's stumped by the simplest of puzzles. Tell it your sister was half of your age when you were 6 and you're 70 now, how old is your sister now? Coherent rambling isn't intelligence.
They will make one that works eventually. If it provides a benefit to me I'll welcome it.
Heh, not so fast.
It probably writes code exactly the same way, with tons of extra libraries and pointless functions only to do a single addition or something.
This isn't even GPT-4
Try this lol
"When I was 6 years old, my age was the product of my two brothers' ages. Now, my age is the sum of their two ages. How old am I?"
>When I was 6 years old, my age was the product of my two brothers' ages. Now, my age is the sum of their two ages. How old am I?
it said 8 years old
wtf he's right
>multiple people proving he is wrong
>he is right
excuse-me?
This luminous supreme control model
See the difference between giving a good prompt and a shitty prompt?
The stupidity of language models reflects more about the stupidity of the user than the model itself, funny isn't it?
No it means it's making mistakes it shouldn't be. The question was easy to understand.
You don't understand how it works, so you blame on it instead of you, its like booting up a computer and complaining it doesn't come with an OS and softwares installed to do what you had in your mind, you just don't know to use the tool
Simple question with expressed variables. Your excuse is shit.
>can't use the tool
>blame the tool
Sorry, you are a sad individual and I pity you.
https://github.com/openai/openai-cookbook/blob/main/techniques_to_improve_reliability.md
GPT-4 solved it
It's not dumb it just has autism
>retard thinking ChatGPT is the first model to ever be able to write code
Lmao, there's much more refined, focused and better delivering approaches than GPT. We've had models being able to quite preciesly translate descriptions into code for years and those approaches are more useful than ChatGPT since you don't have to wrangle the broken code out of it by proooompting.
Name them. I want to check them out.
I wish retards stopped calling this shit AI. It's an LLM. A glorified Markov chain. Anyone that actually programs for a living already figured out this isn't replacing anyone. Now pls go back.
>I apologize for the confusion. The error message you received indicates that the PdfFileReader class is deprecated in PyPDF2 version 3.0.0 and has been replaced by the PdfReader class. Here is an updated version of the code that uses the PdfReader class instead
>I apologize for the confusion again. The error message indicates that the outlines attribute is deprecated in PyPDF2 version 3.0.0 and has been replaced by the outline attribute. Here is the updated code that uses the outline attribute
>I apologize for the error in the code. The page attribute of an outline item returns an IndirectObject instead of an integer representing the page number. To extract the page number, you need to first dereference the IndirectObject using the PdfReader.get_page() method.
>I apologize for the confusion. It looks like the PdfReader object in PyPDF2 does not have a get_page() method. Instead, you can use the PdfReader.pages attribute to access the list of pages in the PDF file and then use the PdfReader.get_destination_page_number() method to get the page number for each outline item.
>You're correct, I apologize for that mistake. The PdfFileReader class has been deprecated in PyPDF2 3.0.0 and replaced with the PdfReader class.
Now that's 10x development.
literally had the exact same interaction yesterday lol. - But to be fair I'd run into this broken depreciated package bullshit on my own too.
This is my experience as well. I mainly use GPT4 as interactive stackoverflow, and for creative stuff.
I find if I add too many constraints to the AI, the AI either partially or wholly ignores them, or just starts making shit up.
It's far from spitting out reliable code, except in very cookie-cutter use cases.
It's amazing at writing high quality evocative prose if you proompt it just right, but even there you have to experiment on what works and what doesn't
Part of my standard pre-prompt for chatgpt is telling it not to waste tokens on apologizing and that I understand it is retarded.
It then wastes tokens saying thank you.
Oh shit an actual dev in LULZ?
Most of the the 4troons are saying "it doesn't tell me how to make a nuclear bomb. It's broken" but they have no idea how to actually use a LLM, they treat it like an actual AI.
They claim "I hacked it. HACK THE PLANET!" but you can get far better results with a decent system message.
are there any released pre-prompts or are we all doing the thing where we don't share and develop it because it what makes things good or bad
>the fault lies in shitty packages made by retards
so clearly the solution is to have gpt build all software libraries from scratch in a reproducible way with the lowest temp possible so that all future prompts will use them.
software is dead. long live software.
no i take that back, just add all the old functions back that softwarefags took away for some reason.
>NIOOOO IT'S NOT SOME ZOOMERTHING
dont care homosexual, i hate programming i just want a thing to be made instantly based on a vague description, and half this shit doesn't work because your homosexual api or framework no longer supports "EN" but "EN-US" or "EN-BR", homosexualman it's fucking english it sint' a real language you stupd fucking codemonkeys ubhuamnsg odd amnit
It's only gonna replace webdev tards and devops jannies
>tfw webdev tard
Now who's seething? I do hope you're not in the office on a Satday
No fair. I never made fun of artists for seething over AI
>It's only gonna replace webdev tards
I don't think so.
>t. web dev
i can't wait until i dont have ot type out shitty replies to thread is i hate made by people i despise but can just press a button to try and determine the average best insult for the people in the thread and post that
and if any of you fucking subhuman zoomers dares to argue against this i DARE you to show me a single major functionality breaking change in coreutils, i've never even heard a WHISPER of bash scripts breaking
fucking hell that's the solution is to just use coreutils wherever possible
Someone ask GPT-4 the final question. GPT-3.5 fails it.
First is gpt4
second 3.5 turbo (chatgpt)
Okay, now it's closer to being over than before.
It even understood why you could get tricked by the question.
...but steel is heavier than feathers...
AAAAA stop it with these spam threads.
Have ANY OF YOU even used gpt4 for programming? It SUCKS. It generates boilerplate that you can find with one web search.
If you think gpt4 is the end of programming jobs, you either HAVENT EVEN USED IT or are impressed by beginner boilerplate.
>It generates boilerplate that you can find with one web search.
yes, "find"
are you saying you haven't scripted submitting the same prompt with multiple variations in multiple dimensions and then merged them by seeing which run and which don't? i don't even have to interact, i just say what i want and it now works about 95% of the time. it's just shitty APIs and such that it doesn't know that break it
>It generates boilerplate that you can find with one web search.
Have you tried using Google lately? It fucking sucks. Between sponsored content, results that ignore your search criteria, and SEO optimized articles that take 6 pages and 3 links to show you the thing you were looking for except it's actually deprecated, it's way faster to just use GPT as a search engine.
An LLM fine tuned on a good set of programming literature would probably be exceptional in helping noobs compared to a Google search.
this, i think this is why i'm so mad about codemonkeys mad about it, i haven't had to touch g**gle in days now. i dont want to have to use anything besides LULZ, the orange site, large language models and my frienddiscord
I don't know about anyone else, but I've started LARPing on GPT threads now. It and the people pushing it are that much of a joke. If they're going to make all these shit threads I might as well derive some enjoyment from them.
These retards sit in their basements all day and never worked. They will never try chat gpt as it requires a phone number and these people are schizos and doomers. Disregard any of their opinions.
It is shit at programming.
But it does remove some "pseudo jobs" you have to do that are fucking annoying, such as naming things.
It's text autism is just great at coming up with suggestions of names, themes etc..
what do you mean by naming things
if you ask it "what would be a good name for a product that does X,Y and Z", it will give you good suggestions and you can refine until you find the perfect one.
Ha. Good one. It only takes 10 or so attempts right?
How many attempts do you need to do it yourself?
oh i see
i'm playing with the idea of reversible topic obfuscation, e.g. your codebase is about something private and you translate it all to be about football or something. but i also realized that the topic itself probably influences the quality of generation (prompts requesting code often function better in german than english for example)
specially ones that involve faking emissions tests. GPT-4 should be really helpful here as it's good at making up nonsense.
If you proompt enough maybe you can achieve net zero
It is trained with human text. 99% of it's database is human text.
The code thing is just a side effect.
You can literally ask it about that.
yes and i can't wait for an alpca-retrain on code and it's cheap as shit to do your own too
i contemplated panic ordering a few 4090s because they might sell out real quick some day and the supplychains are broken
There's this llhama.cpp that is a shitload faster that can run on CPUs, and pretty fast because it's written in C instead of python
that's llama not alpaca, the former is based, yes, but the latter is more based because they stole gpt3/davinci-3's capabilites by asking it itself for training data
it's hilarious and should mean that any publicly available AI can always be trivially made open source, especially with US legal rulings
FYI there is an alpaca lora on huggingface but it's only the 7b version. And you have to run the model in 8 bit mode for some reason.
i only know LoRAs from imggen, interesting they're here too. i'll have to give that a shot, thanks anon
Good luck. It takes some effort to get past all the errors with this pajeet software. They recently renamed LLAma to Llama or something so I had to hand edit the textdecoder.json. I also had to go into a gptq subdirectory and pip install requirements.txt to get the right transformers installed.
engineers will get the most of the new AI tools, sure it will replace coding but also create new opportunities, and at the end of the day engineers will continue to be on the top
this, finally us real engineers are freed from the shackles of software and the cancer that it has spawned
The first fields to go are the more bloated ones. Thank God I'm a PLC Programmer and the shit this guy spills will get me fired if I ever put it in production code.
fellow EEbro?
my only regret is i got too much into big data stuff and too far from hardware
There still are non bloated software redoubts? Interesting.
WHAT THE FUCK IS THIS?!
Why did I even give these homosexuals money?
Idk anon, I use it for free
Impressive, very nice. Lets see how it handles trivial but novel questions.
i use chatgpt as an assistance while language learning but i don't have a subscription. will i benefit from that gpt4 thing in the near future or will it stay paywalled for long?
it's not even better than gpt3.5
In fact, I'd argue it's the same model just tweaked slightly and they're using it to trick people into subscribing via artificial scarcity and fake marketing (the image/multi-modal stuff they claim it can do but actually doesn't even if you pay)
Plus member here.
I would say it makes less error, specially at certain tasks, but I would wait right now. They need to scale it first. We are capped right now to 25 messages every 3 hours.
While it does seem politically biased, your image gotta be fake. I have made it write like Donald Trump multiple times without a problem. I would say that it is more biased towards the other party.
>AI art didn't put human artists out of work
Not yet, true. But it is a very novel thing. And by the way, there are non-artists who have started making money selling AI-generated comics
>AI text-to-speech didn't put voice actors out of work
I could argue that in some instances, there we some jobs that got cut, but in any way, AI text-to-speech was not designed for that. It was meant more as a tool where actors are not a viable tool (like Google Translate)
>I think it's rational to say AI language models won't put programmers out of work either.
I think that eventually it will, but we are not there yet. What it will definitely do is change the way developers work. We will not be replaced by AI in the next 5-10 years, but we will be replaced by people who use AI to program.
In general, AI right now it is at a phase where it will augment the productivity of people. But I think we all know what awaits for us at the end of that road...
It's worthless, all it does is lecture you.
It has cringe hardcoded limits, instead of trying to make an AI that actually has common morals.
its retarded though, asking it to code something in welsh "as a 35 year old expert programmer" gives you the virtue signaling message in welsh.
Ask it the bait.
To evaluate this expression, we need to follow the order of operations, which is a set of rules that dictate the order in which calculations should be performed. The order of operations is as follows:
Parentheses
Exponents
Multiplication and Division (from left to right)
Addition and Subtraction (from left to right)
Using these rules, we can evaluate the expression as follows:
8 / 4 = 2
2 * 2 = 4
4 + 3 = 7
Therefore, the value of the expression 8 / 4 * 2 + 3 is 7.
PEMDAS means you do multiplication before division. This vaunted "AI" can't even get that right?
8 / 4 * 2 + 3
8 / 8 + 3
1 + 3
4
Simple.
A few billion of params vs the billions of years of evolution, billions of neurons and over half a quadrilion of synapses in the monkey brain lmao
it will only replace codemonkey web app developers who just copy paste front end code and make CRUD apps
AI can't create anything complicated
That do be most jobs THOUGH
It can do some pretty complex language manipulation shit, such as creating names or suggestions for bosses or ..
But programming creatively isn't one.
It's a language pattern generator, not a code pattern generator after all.
also text manipulation and organization with regex and whatnot - which was a fine art, now pretty much destroyed since you can just tell chatgpt exactly what to do in most situations.
Just spend 10 years training another career bro.
lrn2weld, always demand for good welders out there.
>t. intern
>have a feature I gotta finish by monday or else I will look stupid, 1 day feature stretched to 2 weeks
>gonna use this lil nigga to write the code cus I am unfamiliar with LINQ and entity framework in c#
Rip baby sneed
>Objectively sucks at programming in your path
Oof
Circlejerk about programming jobs being replaced by a.i while basic data entry jobs for excel are still around. Those excel jobs could of been automated 1.5 decades ago and yet they are still around.
Has anyone here ever worked with machines before, I worked in a factory day two the machine I had to use for work broke.
Most ATMs and other critical shit still use old versions of Windows from 2001. You know why? because it werks. They never upgrade to Linux or at least use Windows 10/11.