AHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAH
AHAHAHAHAHAHAHAHAH
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
>another shitty <80 IQ twitter screencap thread without context
>"let me guess, it's about AI"
>it's about AI
>another seething baseddev
AHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHA
Im an outsider and a layman but from what I've seen posts like these seem like cope, cognitive dissonance and time wasting. The writing is on the wall and I just made my machinist friend shit his pants when he asked me to have it make him this chart. Is there really a point to beating around the bush on this?
>chatgpt makes shit up on the spot
>anons machinist friend trust anon
>his plane falls out the sky 6 years later killing 400 people
I dont mean to be mean but you guys are seriously coming off as boomers. This thing is improving at an exponential rate.
>This thing is improving at an exponential rate.
Goalposts: moved
>heh nice try kid I meant the version now not the one that will be 100x more powerful in 8 months better luck next time
I mean can I at least get an argument as to how it wont take everyones job in 2 years? Or are people here not allowed to think of logical outcomes?
It took 3 years between GPT-3 and GPT-4. GPT-4 isn't significantly better than GPT-3.
>Yes, all the research being worked on, all the papers out now showing massive room for growth, all the domain specific potential left to apply to new models, all the scaling laws, all the new hardware and businesses appearing, all of it will disappear
>Why? Look at this funny dog!
Yeah you know what is realistic, having GPT-5 run on 20TB of VRAM. I don't think you realize this, but when you hit a certain size the shit don't scale.
Anon, GPT-4 is very likely still 175B
Anon I don't know if you realize this but GPT-4 is limited to 100 messages per 4 hours for people PAYING MONEY. And the API fees they listed are highway robbery. I don't know if you realize this, but GPT-4 is already too expensive, too rate limited, and too slow for production applications.
It took 3 years for GPT-4 to come out from GPT-3 and it's marginally better.
It took GPT-4 6 months to come out after GPT-3-turbo which came out a bit after GPT3.5
The fees exist because they want to make money, inference is cheap, even for T+ models.
You bet. And it;s paper"s", you can see scaling and emergent behavior in almost anything published involving a model at different scales. This is one of the most obvious and well known facts of machine learning. We have results into the trillions of parameters.
The rate limiting says otherwise. You're just coping and you're straight up lying about the timelines now. GPT-3, the last MAJOR VERSION, was released early 2020. GPT-4 is not much better than GPT-3 and the only thing that makes GPT-4 better is OpenAI figured out they can fine-tune the model for instructions.
>The rate limiting says otherwise.
The rate limiting says they want to make money.
> GPT-4 is not much better than GPT-3
Ahahahah, shows you have no idea what you're talking about. Anyone who used GPT 3 on release knows it's MASSIVE update. It's less of an update over the refined divinici's but that's why I brought them up.
You being a homosexual and not know what the timelines where doesn't make me a liar.
The jump between GPT-2 and GPT-3 is massive, the jump between GPT-3 and GPT-4 is not massive. Try to keep up. We're already seeing the plateau.
> the jump between GPT-3 and GPT-4 is not massive
Again, you are wrong. Which is funny, because the number of parameters between GPT-4 and GPT-3 haven't changed, but we STILL see incredibly large improvements. Scaling could bring it to the next level, but it might not even be necessary going forward.
I used GPT-3 back when it was released (lots of AI storytelling stuff), and 3.5 blows it out of the water completely. It's not even a contest.
You don't know the number of parameters of GPT-4 because they have announced them. So CITATION NEEDED.
if the model gets bigger, doesn't it cost more to the companies to generate results?
Sure, but it probably isn't any bigger. It's almost certainly based on GPT-3's architecture (or maybe GPT-Turbo's?) but using flash attention and other tricks to expand the context efficiently. Plus the inclusion of an auto-encoder for vision of course.
>They threw out all progress they had and decided to spend another billion training a brand new model despite directly developing and implementing methods to expand it, and despite every piece of published research saying the old model was drastically under-trained and could easily increase performance just by increase the amount of training; that ignoring the easy gains from making it multimodal.
ok
Perhaps you can explain then how do you expand the context window by a factor of 2x for the 8k and 8x for the 32K context model using the old GPT-3 sized model.
I guess you can warm start training using the old GPT-3 weights padded by new untrained weights. But it's unclear how much that would help.
How much is OAI paying you?
GPT 4 is not much better than 3.5
it adds law homosexualry and equations
Also,
>but when you hit a certain size the shit don't scale.
Literally every paper in the last 9 years says you are wrong
>paper
Is it peer reviewed though?
Are CS papers not peer reviewed?
I expect less from CS homosexuals to begin with but i don't believe that.
Peer review is a meme
Yes lol it's such a meme. That's why declined a couple of papers doing review because they where hilariously bad.
>That's why declined a couple of papers doing review because they where hilariously bad
What?
There have been experiments with papers with errors and most of the reviewers didn't catch them.
>what?
Yes anon. I finished my PhD in a field that is rather special Elsevier or Wiley ask scientists who have an expertise in a certain field to do review for papers that fall into that category. And yes I actually rejected papers because they where errors or parts that where not scientific enough. Obviously as humans are doing review there might be cases where errors slip through but should everyone and their mom be allowed to publish shit with source "I've seen it in my dreams".
Nevertheless I also believe that there are too many papers getting shit out only because institutes want to grab funding.
So much this kind stranger. Peer reviews are extremely necessary for OUR democracy.
And citations! Oh my if it doesn't have citations it doesn't have upvotes simple as
Yes, they are.
I thought everyone on this board had some flavor of CS master's degree. Have you never tried getting published? It's a pain in the ass, literal rng as to what reviewers you get
can AI draw hands yet?
This, you have to prove it its right all the time, you should be better making programs to calculate things instead
The problem is you can't trust that at all, you can't trust a single number the AI writes. I wouldn't do any work based on the numbers ChatGPT quotes. You have to be an industry expert to use ChatGPT responsibly and I can't wait until some engineer blindly trusts ChatGPT and kills 1000 people.
Again this is boomer talk.
Hey child how about you use ChatGPT for a couple hours and realize how much it fucks up. It's easy to clap like a seal because you have low expectations but how about you take a second and try to make a personal website or something and find out how accurate it really is.
lmfao ok boomer on Twitter people already turned napkins into websites
Hey child they made a landing page that you can create in 2 minutes using Squarespace. I'm sure a child like you heard of that site before.
>Squarespace
Ok Linus Tech Tips LOL LMAO EVEN
Say bye to your overpaid WFH job. Yes I would like fries with that :^)
As opposed to what? Being some 16 year old who is going to kill himself because he's chronically online on LULZ? Some loner with no friends? Do yourself a favor and ask your parents for some help.
My sauce, wagie, put it in.
Are you over 40? I ask because it would explain why your reluctant to accept this.
Accept what? Change? Things change all the time and if you think everyone's going to be unemployed I'm going to let you on a secret: they're not. Even if AI was a disruptor it's going to take DECADES for it to actually retool the economy. Think about fucking long it took computers to take root. And think about how efficient computers are compared to fucking DOING EVERYTHING BY HAND.
Sometimes I think you children have been completely lobotomized and lack any and all critical thinking skills.
Boomers are unironically the most excited about ai. They think they will all get rich from pushing a button.
>Twitter people already turned napkins into websites
kek yeah the most simple design ever, I am not a coder, but how is that useful?
wait till you learn coding is just a bunch of pajeets, naggers, homosexuals, and women being paid 6 figures in an adult daycare doing nothing but cranking out shit like that and gluing it together
cant wait see to more crying tik thots laid off and replaced by GPT-4 HAHAHA
I don't know if you realize this but most applications are longer than even 35,000 tokens and at 35,000 tokens it's $2 a gacha pull in GPT-4. I don't know if you realize this, but most programming projects have an architecture that spans across 1000s of lines of code across multiple files. Neither of which GPT-4 handles.
cant wait to pull up at McDonald's and see your NEET ass afraid to make eye contact as you hand me my bag HAHAHAHA
I can't wait for Digital IDs so people like you never post again.
tick tock wagie earn while you still can
What kind of projection is this? I can only assume you're underaged and currently unemployed. I can only assume your chronic LULZ use has completely fried your brain and given you depression and nihilism so maybe I'll be working in McDonald's in 5 years but you'll probably have killed yourself. I can only pray that I can find at a job that doesn't make me stressed out as much. You think I would hate working at McDonald's over the sleepless nights my current job gives me sometimes?
But I'm not that lucky. I already have 100gs saved up, my job isn't at risk and if anything I expect major growth in the next 5 years especially since my company won't need to hire more developers they'll just pay me more and more. I already have a house. I don't need much of an income to stay afloat indefinitely. I look forward to the future and unlike you acting like a petty child throwing a tantrum because YOU are the one who is scared, I'm in the process of training my own Stable Diffusion model as we speak.
>they'll just pay me more and more
that's not how companies work dude
It is when you're a partner in a small business.
For perspective in the last 3 months I've gotten $20k in bonuses.
mucho texto didnt read
my bag feels cold
a lot of companies dont want to give away their code to chatgpt either
kek
>1960
>damn this internet thing is shit, can transfer data only at 2kbps
>literally useless trash
You are an underage zoomie, right?
you are the one underage here if you think things will stay as now and not become more powerful/better, if today a thing is 100 in the future will become 200 until it reaches a bottleneck and something better is invented. Saying is useless because "it can't write 100k lines of code today", is stupid.
But either way apart from the processing powere one of it's limits is the quality of the database.
Think about it like a really good archive search, that can bring you exactly the paragraphs from the books you need and help you in 90% of the work, then you do the other 10% yourself.
Anon, you don't know jack shit about history, much less about computer history
This isn't the first time there's insane hype around AI and at this point it seems they're hitting overpromises and a ceiling on the current paradigm, GPT-4 has all the markings of bringing in another AI winter
You mean like VR? You are a one dumb homosexual.
First of all, your code doesn't need to span across multiple files. It's what us humans with extremely limited memory capacity find convenient, but it's strictly not necessary and even reduces computational efficiency of the entire thing marginally.
Whoosh you don't even understand why all of it is a problem because you're just some child who doesn't even code.
>wait till you learn my cope
It literally says in that post
>I'm sorry for my previous output
Meaning that anon had to ask twice (at least) and the first time it was either wrong or gave a garbage output.
>well I'll be damned.. this chatGPS thingermabob gave me an answer based on my question of simple unit conversion... this thing is gonna TAKE ALL URR JOBSS.. WAKE UP SHEEPL
Now tell your machinist friend to make it write up a simple CNC program in G and M Code. He'll probably be impressed, but he'll see the gimmick that it is. Because it's G-Code programming leaves a whole lot to be desired
nagger you're not supposed to make G-code by hand, that's like building websites by writing raw html. You get your CAD to make STEP files and then feed that to CAM which produces G-code.
>advanced calculator is good at advanced calculations, whoda' thunk?
Realtalk though, it's going to put a chunk of people out of work. But I dislike it being called 'AI' it should be called something more like 'Interactive Computing' or something gay. It doesn't think for itself. It's just a fucking advanced response bot, combined to a computer.
its not. The AI currently gets a lot of stuff wrong, and it always converses due to how its programmed in specific models like it knows what its saying, or it is "convinced it's right" when sometimes the errors are egregously wrong. You can test it yourself. It needs a lot of work at this current phase. Will it put some people out of jobs? Yes. It's still about 5+ years off though for any real large scale utilization. There are some jobs though that should start specializing and getting the hell into something else. (Radiology, I'm looking at you.)
>advanced calculations,
It's a division, how is that advance?
If you think AI can just do division, you haven't seen shit, it can absolutely do extremely advanced calculations on the fly. It's going to be an amazing tool for many jobs.
i am talking about what the user posted dude
what's so impressive about simple unit conversion?
its fucking alttroon personal gpt shilling here
You are irrelevant
This must be what all the artists are doing with their time now.
Last night I googled `girl with huge tits and ass`
A (very good) onlyfans thot quoted me $100 for the same racy pictures
Google delivered the same in 1 second, for $0.00
Genuinely mind boggling
lel
gpt cant write C
Not yet
thanks to you assholes that keep feeding the GPT machines
please try feeding my dll to gpt
>when your AI robot gives you a dllhole an indian rugburn because it fucked your ass too fast
I just had ChatGPT write C. Worked great
How much C
>almost 200 lines!
top lel mate
Who are you quoting stupid frog poster
gpt4 can't write more than 200 lines of C at a time
go ahead and try
reality comes at you fast kiddo
So instead of 2 weeks, it took him 3 hours - that's pretty sweet. On the other hand, he doesn't appear to be a (very good) dev, so code quality is almost certainly worse and testing / debugging will probably be somewhere around the level of "trust me, bro."
Of course, these are just the early days of AI-assisted coding. Will be interesting to see what his (very good) dev will be able to quote once he has added AI assistance to his workflow.
Can it write Nim? Should be much easier, but unfortunately there are way fewer examples for an AI to learn from.
Oh really? is that the reason for the hate threads? Is someone gonna call me a cnile now? We laugh at you when you do that by the way. You're a joke.
You literally spammed this a few hours ago with the same pic; fucking stop shitting this board you literal nincompoop
Is it some kind of brigage? I doubt openai has part of their marketing budget allocated to shilling on LULZ, who actually makes these threads?
*brigade
its retards spamming them. They think a normal person would be able to replace programmers when in reality its fucking programmers themselves using the program to make it easy for them.
If that guy didnt know how to code microservices theres literally no fucking way he can tell the AI to code it.
the same with translating text, let's say you want to translate something into German, using software can help you, but if you know 0 German, you would have no idea if the output is right or wrong
What budget do they need when they can just have their product shill itself?
Artists
>Smart person leverages position for money and leisure pace.
Most projects are sub 20 hours when done by a competent dev motivated to deliver.
microshitters BTFO
Why does he attribute the result of his work to the large language model? A lot of people don't know what microservices are, give them the task and the neural network and they still wouldn't complete it, hence it's him doing it, not the large language model. He is using the large language model as a tool. Why not post the same message but saying that his laptop created the microservices? Or that node.js created the microservices?
laptop and node.js didn't just spring into being yesterday.
Any brand new time-saving tool would have people tweet/blog about it.
>things that never happened
fantastic
until these guys post the raw outputted code I'm not interested in what they have to say. society should be better than this.
oh no think of all the pajeets working in big tech
Is there an open source project that has used GPT to produce sizable amounts of code. Tired of screencaps with no code.
snake game
OP is probably a poorfag wagie who had the opportunity to make six figures being a programmer, and now he's bitter he squandered his only opportunity out of poverty.
>So I asked a contractor to come in a put up a new wall and run some wire, he quoted me 1000 dollars!
LOL FUCK THAT
>I asked my wife's son and his friends to put up some plywood paint it and run an extention cord, they'd do it for 25 bucks and pizza!
NOW THIS IS UNREAL DEALS
A lot of this, GPT-4 can do some cool things but you want it secured and hardened? Integrate it logically into other applications? Monitor and troubleshoot when it doesn't work? You're fucked
Answer should be "Not yet" instead of No
Just like self driving cars will be replacing truck drivers right?
the only reason that self driving cars arent replacing truck drivers yet is human safety.
I guess that seems to be the case for code too: You need to ensure that the code written actually doesnt mess stuff up, you cant just let it code around randomly. It has to be very exact or led around by a human (similar to a dude keeping his hands on the steering wheel and checking everything).
At that point the human is still 95% required though and it kinda defeats the point of having a self driving car because you still have to pay that dude. Essentially, for now AI is still an autocomplete tool. Even for driving.
>the only reason that self driving cars arent replacing truck drivers yet is human safety.
who's to say legislation won't be passed in the coming years to protect humans from the effects of ai? surely there will be regulation eventually, especially if this ai shit becomes even more sophisticated than it already is. how long would such legislation last? god knows. the black pill in all of this is that countries that don't adopt ai will fall behind as other countries adopt it. we're almost forced to adopt it to keep up with our adversaries. stopping ai would unironically require a world government.
>the black pill in all of this is that countries that don't adopt ai will fall behind as other countries adopt it. we're almost forced to adopt it to keep up with our adversaries. stopping ai would unironically require a world government.
Fuck I haven't even considered this. We truly are fucked holy shit.
pretty much
>who's to say legislation won't be passed in the coming years to protect humans from the effects of ai?
I think what's more likely are expressways that are dedicated to AI-driven freight traffic and prohibited to human-driven traffic. Similar to streetcar lanes, rapid transit bus routes, etc. Just throw up some warning signs and fences, let commerce flow. At least that way whenever an AI-driven freight vehicle hits someone, that someone shouldn't have been there to begin with. Nobody blames the conductor when a pedestrian is hit by a train, after all.
>the only reason that self driving cars arent replacing truck drivers yet is human safety.
And the only reason GPT/AI isn't replacing coders outside helpdesk Tier 1 "what is your problem?" shit is because AI is so immature and easy to trick into fucking itself over
Actually, the only reason is that they're too shit to do so.
>the only reason that self driving cars arent replacing truck drivers yet is human safety.
Actually it's because they'd destroy close to a million jobs in the USA alone. So they'll never be legalized for trucking, or if they are it'll be done in a very gradual manner where there are quota licenses to use self-driving trucks given out as the pool of truckers decreases with retirements.
I rode in a Tesla that could change lanes by itself it was pretty cool
I believe context length for GPT-4 is 25k words or something. So yes, it can easily put that code in a legacy application with thousands of lines of code.
>Can it test that feature
Yes.
>Can it communicate with backend enginbeers
Yes.
>Can it communicate with managers
Yes.
>Can it test?
Yes.
>But can it... TEST?
YES IT CAN YOU REATRDED APE HOW DENSE ARE YOU
>Can it grab a physical computer
No, but why would it?
>casn it fix an IDE?
lmaooooo
whoever created that pic is literally retarded
The thing is there are a lot of intelligent people that can sort through problems that don't necceserily know how to code, with shit like GPT-4 it basically unlocks a whole lot of options.
It's extremely simple and basic, but I got an analogue clock gui working in python by using AI (Not even GPT, you.com), I have no python experience and only some really shitty C# and java experience from a college course I took when I was 17.
It had a few problems but I just said it wasn't working, and described the errors, after a few minutes I had it. Meanwhile if I actually learned to code in python it would probably take a lot longer. Sure you can't do everything with it, and it's not like a shitty clock matters, but the shit it can do is better than the previous no capability that people have.
And often like OP's pic non-programmers don't know how hard or easy something is they just accept the price and pay.
it's never going to be a problem. normal people don't know how to talk to computers. half the job is figuring out what the business rules are (what the client actually wants). and that's difficult enough when it's two people talking to each other (the client and the programmer). anyone that has worked in the business knows this.
imagine a client trying to explain to a machine what they want the software to do. the AI, if it's smart for real, would kill itself after a couple of hours.
Absolute cope
are you coping?
I guarantee you that 90% of that will be able to be done within the next year
Your guarantee means nothing.
AI can
that's how blackberry has this AI they use for security shit that can find patterns of malware or even bug in large codebase
what microservices did he have GPT-4 right? What do they do? What business logic do they encapsulate? How do they communicate with each other?
>microservices
oh so bloat retard just used AI to do more bloat retard stuff, just shows AI will enable retards to do stupid things
It's over
This guy graduated college in 2016 and his only tech work experience is literally freelancing outside of being a “founder”. What does he know about building systems?
He also said all the scrips were right the first time, which sounds like bullshit to me.
They probably didn't have any squiggles in the IDE so that was good enough for him. I bet if you asked GPT-4 to do a security and performance audit of that code it would have a million things to list.
Bangor is a shitty university, no wonder
Funny because when I go to push the buttons of some brand new scammy fad where some groomed nagger in front of a computer is talking about how they made so much money just doing "this" the opportunity suddenly isn't there but they ask for some identification and cards.
HOLY COPE THE THREAD
do you remember a year ago on how Ai can't do art? Now is the same, you guys are coping in the same way, the only difference is that artist have a following, can create with ai and have their "uniqueness", moreover is a shitty job and 95% do it for hobby, instead codemonkeys are almost all wagies. Get the rope homosexuals, six figures are no more, fight for scarps because only one in 100 will have the honor to be the janny of AI
Every code created by GPT models, has been just copying the most basic github examples like pong, simple programming exercises or answering leetcode examples that have thousands of similar answers on github, plus consulting stackoverflow documentation.
yes and in the same way it does pictures and in the same way thousands of codemonkeys program, remove their interner connection to stackoverflow and millions would kts, itzs just that a certain point the copypase will be good enough, and you will just need 1 good janny to clean it up, and he will be able to do in 1 day the work of 20 people, now not needed and unemployed, like an artist could finish 20 pieces in a day by touchong up the hands and some small fixes. You will be unemployed and you will like it, companies always try to cost cut
The problem is that retards too dumb to get that 100k paycheck are also too dumb to understand why nobody employed cares about any of this shit outside of treating it as an enhanced search engine. And they're really fucking abnoxious because they're really fucking jealous
Most people's jobs are writing basic boilerplate anyway.
That’s 99% of all programming
Anon we get it, you are salty you are too stupid to be one of us. Next time save time typing and just ask if we want fries with that
enjoy your unemployment
great anon, chatgpt already made lose a 10k job to a fellow code monkey, you are a lucky one that you are ahead of the curve, so they maybe will need 1or 2 of you in the future, chop chop the rest. Artist just draw stuff they like, and use whatever method, you aren't in the same league, being mostly wagies
Here's the thing, the artists are imbeciles so they can't even comprehend how great of a tool Stable Diffusion is. If you were smart at all you'd train a model on your style and then in conjunction with ControlNet create a masterpiece using your model in 10% of the time. As a programmer LLM don't frighten me at all, it's just another thing to help me write more code faster. In fact it was really useful the other day because I had it write a bunch of Objective-C code for me and yes, that code would've cost $10,000 for us to have someone else write (and probably poorly).
>If you were smart at all you'd train a model on your style and then in conjunction with ControlNet create a masterpiece using your model in 10% of the time.
And any middleschooler can train a model off your art and do the same thing so there isn't any point
>As a programmer LLM don't frighten me at all, it's just another thing to help me write more code faster
It's a thing to help pajeets shit out code and replace you (but you're probably a pajeet yourself so who am I kidding)
shut up alttroon
why are you even shilling here? no one here is paying for your service
nor shilling for you
keep spamming your shit generations to people susceptible to them
>do you remember a year ago on how Ai can't do art?
As someone who actually wants to use AI art in gamedev... it can't, to any useful extent, or at least enough to replace artists. I wish it did, but at present it's not good for anything besides landscapes and headshot portraits, and even then only if you don't need an exact composition.
And if you want it to do UI elements, forget it. Stylized terrain textures for large-scale maps? Impossible.
It's good at drawing big titty women with cute faces, it's not useful for anything that would help me.
because you are an artlet, for me that I'm able to prompt with lines instead of words you could make much more, but yeah, it will be still a skill issue at the end, (you) will still be looking for artists since they can do much better than you even if you have the same tools
No, ESL, you aren't able to make the sort of things I've described with AI art. Nobody is, or we would have seen them by now.
like I said, skill issue
Prompt a UI with icons for a fantasy game like Warcraft 3. Make icons for the following skills:
- Dragon flight
- Ice breath
- Fairy dance
- Fireblast
- Rampant growth
why does a new app need 5 microservices
he probably just made some CRUD shit on NodeJS, at least this is going to stop all the homosexuals who are learning to code and keep the industry in check
>no proof
I don't doubt AI will eventually be able to write production code, and I hope that day comes sooner than later.
But ChatGPT's code is literally some copy pasted snippets from stackoverflow or github or answers for textbook problems.
How about asking it do something useful?
>bing was gpt-4 all along
And nobody realized. Nice exponential growth lmao. Posting from my 30GHz CPU
kek if anything google has gotten way worse with AI
What's stopping the dev quoting less money and a shorter time frame. Then using GPT himself and chill the rest of the time?
People being cheapskates is nothing new. Not the first "my brother-in-law can code the website in a weekend for $200 so I won't do your $5000 quote". ChatGPT is the new brother-in-law and I look forward to being the clean up crew.
>twitter thread
>no proof
Cool, tell it to make wine run better
>Soi who couldn't cut it in academia cries about peer review on LULZ
lol
>couldn't cut it in academia
>soi
Pick one and only one
Holy cope
>Verification not required.
About what? Are you actually insisting men in academia are gigachads?
wtf is fernwarme
retard nazi
Idk man that's the image I got from kuroba
>retard nazi
Guilty as charged
Bur let's talk about climate change affect on trans folx or whatever the fuck you're gonna ramble about next
>Still seething
Yep, we got a live one. Must have it a nerve, huh?
>still implying academia life is something to aspire to
There's zero hope for you
Damn, you're right. I'll just have to take my 400K a year salary exiting my PHD and think about how much I wish I could have been a truck driver instead.
Dumb nagger bitch
>my PHD
Yep, we got a live one.
It took you this long to figure out the person arguing for academia was an academic? Wow, I can see why you didn't make it.
You didn't sound soi enough to be an academic tbh
Maybe once you finish your phd you'd transition fully
Keep clinging to that anon, maybe if you call them all trannies long enough it will make your McDonalds wage slaving more palatable! (it won't)
I'll concede defeat if you type "nagger nagger nagger"
I won't just say "nagger nagger nagger"
I'll call you the blackest gorilla nagger to ever nagger on this nagger planet you fucking GIGAnagger
>I'll just have to take my 400K a year salary exiting my PHD
The average salary here in g seems to be 300k kek
Must be a lot of Masters and PHDs then.
some of that and some lying on the internet i am guessing.
feels good to not be a webshitter
should i unironically go to culinary school
>he screenshotted a lie and reposted it on LULZ
Okay retard.
We are safe
>human trait replaced: empathy
>Email marketer
Look at that cunning little shit, pretending like it's harmless so it can put its foot in the door.
How long until AI replaces all 99% of politicians?
Forever. People belong to ruling class does get ousted simply because of its incompetence. Politicians will be safe. Public employees will also be safe. Taxpayers, otherwise, are doomed.
All we need is a sufficient push from the voters to cut government spending and automate jobs of political nomenclature. People would rather serve AI overlords than a bunch of senile fossil fucks siphoning tax money.
>we need is a sufficient push from the voters to cut government spending and automate jobs of political nomenclature.
Liberal democracy is designed to prevent voters from pushing such kind of proposal t h o u g h. It doesn't matter whether voters want it or not. It took literally two centuries to stop politicians from arbitrarily raise their income by 27th amendment.
People back in late 1800s and early 1900s really wanted communism but not a single country (except the Czechoslovakia) achieved that simply by vote.
Russia might do it. They're already cutting massively on the buerocratic machine with the whole gosuslugi thing, which is basically an automated digital government office. This shit takes care of so much crap that you basically only need to visit brick and mortar offices to sign shit or grab your finished papers, and only as a physical proof because the records are all digital anyway.
>Russia might do it.
Not after their "special military operation" t h o u g h. Biden banned export of all the GPU, CPU alike and it will remain in place forever.
Also voters don't make change in Russia.
Nobody voted for installment of Gosuslugi and yet here we are. At this point you'll have a very hard time of even scheduling a government related operation without using one. Russia is legit a real life cyberpunk country.
> Biden banned export of all the GPU, CPU alike
Gee well how come I just bought a 4090 and a 7950X? Shouldn't that have been impossible? At least not at the same price as in other countries, you think? In any case, Russia has its own electronics, it's not the best but it will do the business.
Russia might not even exist by the end of 2024. They also have one of the most inefficient, bloated government structures ever, full of apparatchiks, former kgb/fsb glownaggers and just old fart "yes" men. Most of them don't even know how to use pc. Not to mention that russian cattle is incapable and unwilling to change anything in their country.
Buerocratic inefficiency might've been the core reason why they're pushing so hard for Gosuslugi. Now it's too late to stop it. Bean counters have been long dealt with, now pencil pushers are getting the rope. Soon enough there will be AI mayors and then governors, that I guarantee.
>Russia might not even exist by the end of 2024
So might USA. For what its worth, Russia endures economic damage well, whereas US shits the bed at a slightest upset.
Do you realize that majority of russian population is simply unable to use Gosuslugi due to lack of pc, internet or a smartphone? Do you understand that babushkas don't even know how to use them, even if they could?
Besides, the push for Gosuslugi isn't made to eliminate useless government jobs, but to make fsb glownagger job easier. You still need them to keep cattle in check.
>So might USA. For what its worth, Russia endures economic damage well, whereas US shits the bed at a slightest upset.
You can't be fucking serious.
Do you realize you're so full of shit that you don't even have to talk using your ass specifically, as the only people who don't own a smartphone are the ones who intentionally choose to stick with a dumbphone? I know you w*stoids like to pretend like you're hot shit because in your mind the people you oppose are all inferior by default, but Russia is among the most tech-literate and advanced countries in the world. The whole reason like 90% of inventions and technologies aren't credited to Russia is because it can't pay its scientists a top dollar so they work for other countries which gets the credit ultimately. The amount of shit made by people from Russia is honestly ridiculous.
I think you are a homosexual.
I am of the same opinion about you. How the fuck can you be so clueless about your own country? Unless it's not your country of course.
>Activate Windows watermark
>https://en.wikipedia.org/wiki/Medal_of_the_Order_%22For_Merit_to_the_Fatherland%22
So you are a fucking glownagger. Makes sense why you defend this shithole so hard.
Nah I'm a merc war veteran. Breddy gud money all things considered, a lot of type 3 fun though.
From your language I gather that you're an angsty 13 years old, and for your own sake I hope it's physically not mentally.
>So might USA. For what its worth, Russia endures economic damage well, whereas US shits the bed at a slightest upset.
You might think so. The US recycles their leadership in economic turmoil at the drop of the hat. Having a bad economy is almost insta lose in elections. That is a partial reason why they don't have lasting economic damage.
>They're already cutting massively on the buerocratic machine with the whole gosuslugi thing
Yeah, then you suddenly notice that somehow your digital signature was used in digital deal of selling your property
Going full digital is not foolproof
Avoid full digital as much as you can
You don't have to get a digital signature because it's a fucking USB stick at best. It's obviously not gonna be a requirement until RFID chipping has been normalized.
codebros
it's over
we got replaced faster than retards drawing dicks
Yeah this is bad, the estimate me and my dev coworkers had was 5-8 years before we are obsolete, the pace seem much faster than that.
its actually scary as fuck
you can input any problems in leetcode and it can solve it for you in mere seconds
Because thousands programmers upload solutions of leetcode on github.
You know there's an entire page of solutions for each problem right ?
>you can input any problems in leetcode and it can solve it for you in mere seconds
according to openAI's own data it gets almost every leetcode problem wrong
But does it work??
>another thread of people shitting themselves over Google2: Lazy Zoomers Edition
yawn
Google can't do any of these bro..
AI written code has no soul
i'm using GPT-4 for refactoring some C# classes
its barely faster than doing it myself
it took me 3 hours of copy/pasting to finish 1000 lines
>3 hours to magically generate it from scratch
the token limit contradicts what he's describing
if this guy got GPT-4 to write 10 days of code then that means he knows how to architect it himself and he's doing
>"write this function, it should be like this"
>paste it in
>"write this function, it should be like this"
>paste it in
while also doing code review etc.
fake and gay
OpenAI is just marketing itself
>OUR NEW PRODUCT WILL REPLACE YOUR ENTIRE DEV TEAM
>SUBSCRIBE NOW, AND INVEST NOW BEFORE IT'S TOO LATE!
perpetuating hype etc. with bot shills
You're using more than 32K tokens? Wow! Sounds like more than simple refactoring to me.
>token
>word
these are not the same thing you retard
32k tokens is not that much code
>>word
>these are not the same thing you retard
Never said they were, you feeling alright?
>32k tokens is not that much code
It is. It is several thousand lines on average.
No, it's not. Go try for yourself.
>he thinks code and text are the same
gpt4 can't write more than a few hundred lines of C at a time
thats jack shit
DUDE MICRO SERVICES
This is like me cutting my own hair.
I can do it and get ok haircuts but if I want to actually look good because I can't risk a bad haircut I'll go to a barber I know that is a fucking artist.
More complex projects are like building a house off youtube tutorials.
>LULZ's typical fizzbuzzer feeling threatened by some dumb script
DO NOT REDEEM THE CODE
what the fuck is a Microservice
t..30 year old EE
>HAHAHAHA x1000
This is one of the most cringiest post memes out there
>Learn to cod-ACK
I've seen everyone say that chatgpt can write code and make working projects but I've never seen a single example of something more than a simple poojet example that could easily be found on stackoverflow or github
Because these threads are just shills trying to grow hype around the new tech.
ChatGPT uses enormous resources to autocomplete short snippets of code that often don't even work if you ask anything non-trivial(eg not something you'd typically find on stack overflow or in examples folder). As impressive as GPT is, it isn't scalable and we are pretty much reaching the end of what we can do with it. If you want anything more, you need different approach, but that's not the point here. The point is to get as many investors as they can before they realize there is no more beefy GPUs to put it on.
Can’t wait for people to start proompting entire video games and websites and laughing in your face about how you’re gonna be unemployed.
You reap what you sow, homosexuals.
How are candy crush and flappy bird clones being chruned out at an even faster rate going to make me unemployed?
Two more weeks, trust the plan.
Remember when SD 1.0 couldn’t even do anime?
Only took a month for NAI to come put and suddenly it could do perfect anime.
Exponential progress, poojeet. :^)
>Remember when SD 1.0 couldn’t even do anime?
>Only took a month for NAI to come put and suddenly it could do perfect anime.
Anime is symbolic drawings, it's easier to do compared to realistic style
99% of anime art (at least the kind that tends to get posted on LULZ) is the same generic looking soulless cookie cutter shit. It's the perfect art for AI generation because it already looks computer generated to begin with.
There is literally no technological difference between SD or NAI. They literally just trained it on different image base.
There is so many different examples of actual progress in AI tech in recent 2 years, but yet you've chosen example that shows none of it.
Progress isn't exponential, it follows an S-curve. It only seems exponential at the beginning, then it stalls out when it approaches its fundamental limits. From there you can only upgrade by replacing it with something else entirely.
>t. retard
Bro, this is it. This is the start of AGI. 5 years from now we're all going to look back on posts like this and laugh.
As impressive as it is, at the end of the day it's only a Markov chain. Intelligence requires other things entirely than being able to mimic human speech really well.
>Progress isn't exponential, it follows an S-curve
Source? Progress of what? Did you just draw a shape in a chart and say "bro this is the universe and shit". This post is retarded in so many levels.
The fact that progress on any one specific thing cannot accelerate infinitely is as obvious as the fact that the emission of photons cannot increase infinitely with temperature. Why? Because infinity is not a thing. It tapers off somewhere and grinds to a halt. The speed of technological progress might actually follow a black body spectrum for what it's worth.
yes the curve will start flatting once it has overtaken the universe and is exploiting its resources at a steady rate with better methods only being discovered by its intelligence at increasingly less frequent occasions
lmao seeting ARTfag coping. devs hurt you? don't worry we make sure AI dev will be the last job to lose value hehehe
>still can't replace GFs
now Joe has to do customer support for 5 micro services because GPT-4 doesn't remember what it wrote
LLMs can remember things though?
You're retarded if you're a code monkey right now (99% of you) and not already using it as a co-programmer
If you haven't switched over to cybersecurity in order to exploit all this future code that babies are implementing without knowing its scope, then you're doing it wrong.
>Cybersecurity
Literally one of the jobs AI can easily replace.
Surely the AI will audit its own code. Surely people won't just copy and paste whatever it feeds out. How dense can you be?
You're implying that a human is intrinsically more attentive and less error prone than an AI, which is obviously bunk. Remember those chinks who intentionally put backdoors in Linux just to see how long it takes for people to figure out, and they never did?
The point is that you literally have a program writing code of another program. Not only that but it creates said code off the past inputs of others. This is why ChatGPT and every other prediction AI like it will always be vulnerable and everything they write will be vulnerable. They're called 0days for a reason. Not to mention the sheer scope of trying to have an AI audit an entire network of interconnecting vendors and protocols.
Cybersecurity is a meme to begin with, AI codegen usage rampant or not.
Depends on your definition of cybersecurity. Red/Blue teaming is 100% not a meme I can assure you.
So does it actually run?
Aside from the fact that it imports modules from non-existent fake libraries, it works perfectly fine.
So uh, the plan is for us to all just become neets on UBI when we are obsolete right? Fully Automated Luxury Gay Space Communism right guys? We won't all just be genocide because our usefulness has come to its end right? Hahaha...haha....
The preciousness of human life itself will be the only thing that sets us apart from the machines. Therefore, it will be the only thing with value that we can barter. Bloodsport is back on the menu, we'll be fighting each other in coliseums over what little remains of human-controlled wealth.
c-codemonkey bros we'll be fine r-right?
Maybe. Probably not. At least if companies who hire tens of thousands of programmers keep laying off tens of thousands of people at every hyped up news article's release.
Consider that GPT-4 has essentially been fed the textual and visual (not sure about videos but it seems they also pushed in all images they could find) content of practically the entire internet and it is still not even close to writing anything more than single file projects, I'm pretty sure it will be okay.
>it is still not even close to writing anything more than single file projects, I'm pretty sure it will be okay.
holy mother of all copes: the post
Well go ahead make upload some examples then, shouldn't take long.
post some github repos or gtfo
This, it's improving exponentially, Kurtzweil says they'll pass the Turing test by 2029 but i think we are about the cross that threshold much sooner, in only a few years now, if not a few months.
it's already past the turing test if you let normies at it
even as retarded as it is, it beats out 70% of humanity hands down
Not only that but you can really tell the posters in this thread that *actually are professional programmers* and the trolls. ChatGPT can write your code all you want, you still have to know how to implement it. Navigating the maze of dev technologies is a skill of its own.
everything's fine, axe users
a chainsaw can help you cut trees down all you want, you still need to know how to use it
A chainsaw can also kill you a lot easier than an axe if you fuck around with it without knowing how to properly use it. How many people have delimbed themselves with an axe compared to a chainsaw? And yes the metaphor still holds.
>GPT generates code for 5 microservices!!111!!11
It can't even generate sensible code for one microservice, pic related. Nothing tricky, nothing obscure, the simplest microservice that I personally worked on some time ago and it fucked up literally everything it could.
>second point misses the point of the whole task, it's about subscriptions not channel's uploads
>third point talks about some authenticated user and OAuth which is completely irrelevant here
>talks about Channels.list() which is completely useless here
>claims that you can use Videos.list() to get first 5 videos which is false, you can't do that
>adds useless caching library that is literally just a Map with setTimeout
>sorts videos by `uploadDate` a field that doesn't exists in YT API, it should be `publishedAt`
>sorts dates using localeCompare
>doesn't provide any actual API calls
>doesn't provide any microservice-related code
>completely misses the whole complexity of dealing with YouTube API and the fact that you need to fetch these weird resources like upload playlist, playlist items, etc before you can even get to the videos
>the code is awkwardly cut in the middle
I want OpenAI shills off my board.
that's GPT3 you retard codemonkey but keep ingesting your daily copium
Ok. Ask GPT-4 this:
>Write an NodeJS microservice that using YouTube Data API and a hardcoded list of YouTube channels IDs, it fetches first 5 videos from each channel provided channels are subscribed to and returns them all as one RSS feed. Make sure to not make more API requests than necessary and sort the RSS feed entries by video upload date.
I will wait. I want to be surprised.
>I will wait.
or you could stop being poor and do it now
>that's GPT3 you retard codemonkey but keep ingesting your daily copium
>holy cope
>or you could stop being poor and do it now
Ok, here is your GPT-4. It completely misunderstood the task and just fetched channel's videos instead of subscribed videos which is the hard part and whole point. The instructions it linked do not even have any information how to get the api key, it talks about different authentication method altogether.
I want OpenAI shills off my board.
You think maybe the problem is that you're fucking retarded?
>"subscribed channels" is a per-account private data and the script will have no access to it in any case
>the list of videos is right there except it's not supposed to use it?????
>getting an API key is not relevant to the task but you complain anyway
>>"subscribed channels" is a per-account private data and the script will have no access to it in any case
https://developers.google.com/youtube/v3/docs/subscriptions/list
>>the list of videos is right there except it's not supposed to use it?????
I never said anything about a list of videos
an API key is not relevant to the task but you complain anyway
Yes because it posted nonsense.
>"subscribed channels" is a per-account private data and the script will have no access to it in any case
Not "subscribed channels", but "each channel provided channels are subscribed to". I'll admit the wording is retarded, but I can't come up with anything better.
You can fetch channel's subscriptions just fine unless someone goes to their youtube settings and set their subscriptions as private. They are public by default and accessible from API.
ESL
I meant that you can see channels subscriptions but not channels subscribers.
misunderstood the prompt.
A language model would understand “subscribed to” since to at the end of a statement had fairly specific usage.
GPT3 had this issue as well. It can’t infer the scope of the request so it goes off base the more complex the job.
>It can’t infer the scope of the request so it goes off base the more complex the job.
These really are just souped-up search engines then, huh? Is it really so impossible to make an AI that says "I don't know"? Oldschool search engines like Altavista were better at finding information, at least in the hands of someone who knew boolean. At least with those, no results meant no results. Now, no results means 1,000 pages of "Did you mean this?"
holy cope
Realistically what it may eventually do is become an assistant/verbose API search engine, not actually replace the programmer. For filling the role of the average dev it is just too inconsistent, the output and context too small and error prone, and we're already close to a maxed out dataset. One very limiting issue for GPT and one that has been a major issue in abstractive summarization is factual veracity and like all previous LM it will and can output false statements. Therefore there will always need to be a human to verify the output and I'm not sure that is a solveable issue now or in the near future with just giant transformers.
Yeah, I agree with that. I personally already use AI as an assistant very often. Not for programming, but for story telling(NovelAI). It's very useful, so useful that I sometimes wish I could just press button and autocomplete my LULZ post when I am feeling to lazy to finish it myself. But I also know the limits and quirks of such solutions. ChatGPT is of course great improvement over something like NAI, but it's far away from something that could even do your tasks under your supervision on its own, let alone replace you.
russia is always a backwater shithole that modernizes rapidly when attacked
kek manual labor peasants flooding LULZ with their built up envy and copium
software engineers are in zero danger go back to the elements, back problems and low pay ty
>software engineers are in zero danger
For now.
99% of non-software cubicle jobs will be gone before software dudes lose their jobs
Yes they will be, you are correct. But after that, no matter how long or short it takes, years or decades, perhaps even after your lifetime. eventually A.I. will be able to do white-collar jobs as well. It's not a matter of if but when anon, no one is safe and you are highly arrogant if you think you are, Well i suppose you win if its after your lifetime so if that scenario is the one to happen then fair enough, but your kids are definitely fucked.
feels fucking weird man, ngl. i love ai, use it for cooming on the reg. but it's not going to replace any actual dev. not even close. lot of these schizo neets don't even understand what being a dev entails.
It will replace codemonkeys. Which is great honestly, those fuckwits are 90% responsible for software being as shit as it is, the other 10% blame lying on deadlines being a thing.
it seems like we're reaching the top of the sigmoid curve for transformer based llm
I wonder how much farther it can be refined
I tried to get it to write a HBS case about a small gaming studio winning a government contract and needing ISO9001 compliance
I tried like 6-7 prompts but always got simplistic outputs, I think it just can't do anything even slightly unusual well
>search for code problem
>find tranny
why it's so common codekeks bros?
how higher will go the 41% after this?
>why it's so common codekeks bros?
coders are autistic
trannies are also autistic
there's a lot of overlap
you will always need someone that still knows what they are doing. so highly qualified people are safe for at least 200 years. Do you really think the retarded BlackRock investor wants to proooompt himself? Even if the AI was completely perfect, just prompting "make me a software that does x" wont cut it. It cannot read your mind. Coding is as much about the code anyways.
So yeah instead of coders going homeless, it will be coders using AI to code.
Don't worry, OP. If programming devolves into proompting, I am quitting this industry myself. It's not what I signed up for when I started in the early 90s.
What will you go to do next?
if you started coding in the early 90s and not at least 8 figs yet you're a failure as a human being anyways
That would be nearly impossible scenario where I live. Back then we barely escaped soviet rule (Poland), and average salary was like $20/month or so. I am not complaining tho. I already payed off my mortgage, and my savings would last me like two decades even if I did nothing.
Back then tech was actually exciting. Programming Amiga, DOS etc. I loved tinkering with it on low-level, eventually got good, and landed a job. I wasn't in it for the money, but it also came eventually (of course not Cali levels). For me, tech already went to shit since cloud came, so I am not really that sorry if it dies entirely.
>excuses
that's a long ass way to type "yes i am"
GDP per capita in Poland is $18k. Even six-figure salary is considered sinfully rich in Poland.
This.
Feels good being a richfag while not even try harding in some corpo or whatever.
the cdprojekt guys could do it why couldn't you
>thinks CDPR pays 6-figs to their Polish code monkeys
kek
cdpr was founded by codemonkeys you fucking retard
so you are seriously asking some LULZ anon why didn't he start a multi-million dollar IT company? lmao. why didn't you, retard?
>It's not what I signed up for when I started in the early 90s.
What did you sign up for?
It needs to have some sort of built-in code verification. Asking ChatGPT for programming help is annoying, as it usually goes like this.
>ask for code
>ChatGPT generates code
>code doesn't work, throws an error
>"I'm terribly sorry. I made a mistake. Instead of using function X, use function Y."
>try again. new error.
>"I'm terribly sorry. I made a mistake. Instead of using function Y, use function Z."
This back-and-forth gets annoying fast.
coders are going away KEK TIME FOR A NEW JOB WAGIE
it's interesting seeing the back and forth, the zig zagging of this technology, from obviously replacing tech workers first, to, oops, no, it's replacing artists, back to tech workers
I'm a professional programmer and an amateur artist, and AI art has limitations that I don't see being solved any time soon, even if tasteless retards are impressed by it
meanwhile, I've been applying for management and architectural positions after over a decade of saying no to those opportunities
the writing is on the wall
(I still think you need a programmer's mindset to use these tools effectively for development, but so much of what we do is boilerplate already, I can see a single tech lead replacing a full dev team, maybe using a junior janitor until AI can do it)