You can do the same thing with conditionals in 1 line.
2 months ago
Anonymous
...you mean inline conditionals? that's a better solution? I hope you don't actually interview anybody.
2 months ago
Anonymous
Can you explain why efficient recursion is bad before resorting to ad homs, or is it safe to ignore you at this point?
2 months ago
Anonymous
Not that anon, but you do know that recurssion has added cost for no benefit in this case? Compact code isn’t better code. Performant code that is readable and easily modifiable is good code.
2 months ago
Anonymous
>opinion
i can't do much with this
2 months ago
Anonymous
And neither can anybody do much with your code.
2 months ago
Anonymous
hello sir, welcome to brampton
2 months ago
Anonymous
>now using recursion instead of the loop
it's bad because a loop is simpler and recursion creates a dependency that transcends the syntax (loop actions are contained inside the indented block, but recursion goes back to the function definition).
It makes it harder to write, harder to reason about, harder to change in the future.
The version with if...elif is perfection.
You will take it as an hominem, but I used to be like you when I was 15 and writing 'smart' solutions in haskell, lisp and all that. Overengineering is the main problem with junior devs because they make challenges for themselves out of simple problems. With time you will learn that most times, a 'dumber', more verbose is better because it's incredibly easier to debug it and change something.
Recusion only has a reason to exist when it's an inherent property of the problem. Tail recursion works = use a loop. Problem has a nested repeating structure that can't be done as a loop, without emulating recursion at least (eg. a queue) = recursion probably the best solution.
2 months ago
Anonymous
Although there are valid reasons to emulate recursion sometimes, because default stack limit may be too low and changing it may be impractical. Depends on the language.
2 months ago
Anonymous
>simple recursive statement >transcending syntax >can't parse a basic b***h fizzbuzz in 1 line >needs dozens of them and a loop to do the same
I'm sure the women's code bootcamps love you. As far as I'm concerned, you can leave my office now.
2 months ago
Anonymous
Anyone with actual coding experience just rolls your eyes at you. You will roll your eyes at yourself now in a few years time.
2 months ago
Anonymous
Peak midwit.
2 months ago
Anonymous
You are so smooth brained it hurts. You can leave my office now, anon.
2 months ago
Anonymous
Over engineering is the worst blight on software. Devs who over engineer don’t understand how much damage they are causing when they spend more time than necessary to solve a task in a more complex way than is needed on the off chance that they were able to predict the future accurately. Which they never can. So in best case you get a nuke to blow up a treestump. In the best worst case you have to rewrite everything when new requirements come in meaning that their overengineered solution is actively detrimental. And in the worst worst case, also known as the typical case, the overengineered solution stays and everyone has to engineer around it to solve the actual requirements.
2 months ago
Anonymous
Yes. The rule is simple, code for the current problem >it may need X and Y in the future
no, frick off >it's better to make it more flexible by using an abstraction
no
sadly, this is something that can only be learned by suffering from mistakes
2 months ago
Anonymous
Wise men at least learn by suffering from the mistakes of others.
2 months ago
Anonymous
In my case it was a combination of semi-justified arrogance >be genuinely smarter than almost everyone (3+ std in iq tests, and real life situations requiring intelligence) >treat normal problems as a puzzle >realize only much later when I need to change something I hurt myself once future requirements go beyond the structure I designed before
being the smartest person in a room is bad for everyone involved
2 months ago
Anonymous
Its imperative that your manager not be able to think he understands the work you're doing or your job is in peril
2 months ago
Anonymous
Funny thing is that I’m the manager. But because I’m also a dev at heart I know that I know all the shit they are doing and the best solutions, but also to stay the frick out because I’m not them, I’m using old data and knowledge and I’m not getting paid to code.
I also know how all dynamics changed when I became a manager. People I could talk with before now get the deer in headlights look when I ask innocent questions out of pure curiosity.
2 months ago
Anonymous
Its just the unfortunate reality of work that alongside all the interpersonal functions a manager provides like guidance and perspective they are also the slave driver who decides if the wagies are doing enough work to stay employed so when they ask work type questions people are always defensive. I'm sure all the layoffs aren't helping that dynamic either.
2 months ago
Anonymous
Ask me how I know you're employed
https://i.imgur.com/MAWpEHs.jpg
>simple recursive statement >transcending syntax >can't parse a basic b***h fizzbuzz in 1 line >needs dozens of them and a loop to do the same
I'm sure the women's code bootcamps love you. As far as I'm concerned, you can leave my office now.
class Fizz implements FizzBuzzStrategy {
public boolean matches(int number) {
return number % 3 == 0;
}
public String getOutput() {
return "Fizz";
}
}
class Buzz implements FizzBuzzStrategy {
public boolean matches(int number) {
return number % 5 == 0;
}
public String getOutput() {
return "Buzz";
}
}
class Number implements FizzBuzzStrategy {
private int number;
public Number(int number) {
this.number = number;
}
public boolean matches(int number) {
return true;
}
public String getOutput() {
return Integer.toString(number);
}
}
public class FizzBuzzGame {
private int start;
private int end;
private List<FizzBuzzStrategy> strategies;
public FizzBuzzGame(int start, int end) {
this.start = start;
this.end = end;
this.strategies = Arrays.asList(new Fizz(), new Buzz());
}
public void play() {
for (int i = start; i <= end; i++) {
String result = "";
for (FizzBuzzStrategy strategy : strategies) {
if (strategy.matches(i)) {
result += strategy.getOutput();
}
}
if (result.isEmpty()) {
result = new Number(i).getOutput();
}
System.out.println(result);
}
}
public static void main(String[] args) {
FizzBuzzGame game = new FizzBuzzGame(1, 100);
game.play();
}
}
>meme dependency injection that fricks up the result if you register the Buzz strategy before the Fizz one >handling the FizzBuzz with concatenation after all that OOP fanfare >Number class implements FizzBuzzStrategy to handle the case is which no other strategy matches, but then this case is managed separately making this inheritance unnecessary
Pajeet cargo cult tier.
2 months ago
Anonymous
>cargo cult
Oh my friggin reddit! You grok [0] what "cargo cult" means?! You're so fr*kin cool and smart.
[0] "Grok", a verb coined by Robert A. Heinlein in his 1961 science-fiction novel Stranger in a Strange Land to describe a form of understanding
2 months ago
Anonymous
ty for the book recommendation kind stranger!
2 months ago
Anonymous
>immediate seething reply
It's a funny toy regurgitating silly github code, why aren't you having a laught at the silly code?
2 months ago
Anonymous
The corpocuck style code is silly and stupid but so is your manner of speech. I can think of few terms more quintessentially reddit/hackernews/midwit than "cargo cult." Maybe duck typing and grok.
2 months ago
Anonymous
>I don't like a word you used therefore you're reddit
Tongue my anus, Black person.
2 months ago
Anonymous
I feel like "midwit" is doing a lot of work in that sentence. Do better.
2 months ago
Anonymous
Are you on your period?
2 months ago
Anonymous
There's nothing more reddit than talking about product X and then dropping a "enter product X" for no fricking reason other than being a redditarded reddit Black person.
2 months ago
Anonymous
because what if the silly code isn't silly and i'm just dumb
2 months ago
Anonymous
>>meme dependency injection that fricks up the result if you register the Buzz strategy before the Fizz one
first, there's no dependency injection - it's a hard-coded list in the class where it's used (though I did ask for one in a later question)
second, if you're going to be absurd like that, modification to any implementation will frick up results if you change the order of things unless you also implement explicit ordering/priority
the FizzBuzz with concatenation after all that OOP fanfare
you can ask to refactor that further into OOP, works quite well (did it in a subsequent question)
class implements FizzBuzzStrategy to handle the case is which no other strategy matches, but then this case is managed separately making this inheritance unnecessary
it's actually because I asked for a few back-and-forth changes
copilot's smart enough to do minimal changes to the code rather than completely bulldoze and rewrite everything differently at the smallest change request
previous implementation didn't use concatenation, had strategies for FizzBuzz, Fizz, Buzz and Number, and just checked for the first matching strategy to get the output
I really don't get the idea of this ai stuff taking over software engineering jobs. software engineering requires you to use your reasoning skills and problem solving skills to make software. All these "AI" do is just take testing data and use that to predict what word comes after the next. even in an example like this it just takes the training data and copies it. it even left in the example comment on the fortran section and it even says "welcome to w3 schools" proving that it just parrots what it's given. these are impressive tools but if it keeps operating on the same text prediction model that all other ai use, it's just another gimmicky ai product.
just ignore that. ai doesn't simply scrape content from the web and regurgitate it like some kind of search engine, it's actually learning exactly the same way you and i do. your brain is nothing but a primitive neural network without any adaptable structural hierarchy or recursive phasing. ai is amazing and it's going to take your job chud.
>adaptable structural hierarchy or recursive
Why can't someone just figure out a technique that copies everything a brain can do? It seems like it would be simple
2 months ago
Anonymous
Woah why didn't we think about this before? I'm gonna go try this out in the lab! You're so smart!
2 months ago
Anonymous
I meant, why is it hard? Why doesn't that work or why isn't there a way to do that?
2 months ago
Anonymous
You're so right, it should be easy! We just weren't trying hard enough! It's so obvious!
2 months ago
Anonymous
you're asking why is (human) consciousness hard to recreate, which amounts to why is it hard to create life. moron
2 months ago
Anonymous
I mean, sure. Why would it be hard to recreate (this part of) life?
With how many billions the people of coin have poured into the devil that is AI, this is a laughable improvement, if you can even call this "an improvement" at all.
Is it really downplaying improvements when the shit boomed like six years ago and it can't even replace my local Rajeesh coding for rent money with ten of his benchods yet?
With how many billions the people of coin have poured into the devil that is AI, this is a laughable improvement, if you can even call this "an improvement" at all.
That benchmark is or AI doing SWE tasks unassisted. With human assistance it will be able to do the tasks near 100%. This would mean fewer job opportunities with SWE roles
No, because sadly every time we've invented a way to somehow need less SWEs to do things, we just end up needing more SWEs because less viable projects become more viable.
This is what's happened every time we've massively increased SWE productivity.
2 months ago
Anonymous
>But the number of projects would increase.
Sure but the number of human SWE wont
2 months ago
Anonymous
actually, it will and that is the part which most recognize is the worst.
the iq barrier for entering the SWE domain is becoming lower and lower.
web devs are the Black folk of SWE world, but soon, SWE will become the Black person of engineering, because it will be flooded with single digit "developers", like the web was flooded with "hTmL/CsS engineers".
go to youtube and watch completely tech agnostic people trying to spin up a "10k/mmr saas project".
every single step is a prompt input.
these are subhumans we are talking about..
there are normal, smart engineers, then there are non-individuated NPCs, who usually work at frontend.. but this class now is even below that, they barely qualify as human, only in physical appearance.
SWE is doomed.
question is, how fast will other domains be affected.
this fake AI can be trained to learn how to optimally control those robot hands with which some hospitals perform surgery.
surgery could be completely automated away, potentially.
this is in an ideal future, where the stupid AI does not hallucinate shit and can solve unexpected problems that arise on the fly, due to is emergent "smart" properties.
Just like how high level languages led to fewer jobs than when programming was done in assembly. More productivity = fewer jobs, because the amount of work to do is constant and making programming quicker doesn't increase the demand for programming.
>Just like how high level languages led to fewer jobs than when programming was done in assembly
When high level languages started becoming mainstream the tech sector itself was niche.
AI is becoming mainstream to an already mainstream tech sector with capacity to cause large job displacement. I'm sure AI will create more jobs in the future but the transition is going to be pretty rough
No because Devin is designed to be human-free. This is why Devin crushes all the competition, because its a mouse competing against fish in swimming contest, absolutely not comparable
>>14% accuracy after blatantly training on the test data
By a little known AI startup. OpenAI is gonna up the numbers and other AI players are gonna try to catch them within an year
Thin is yea it'll reach a plateau that's impressive BUT for the most basic things that are spammed across github, for everything novel, curveball problem and so on it'll forever be a random solution generator.
yea good point, Moore's Law died around 2015, they already knew by late 1990s how milk the last bit of possible performance not wiht gate size but with its height lol solely with density grind which again it kind of already reached its limit
For real. People forget the industrial revolution was built off the back of essentially a single thing: converting stored energy to automatic movement. Sure, we hit the top of the curve eventually, but it took almost a hundred years and utterly transformed humanity.
How will you midwits cope the moment a model achieves the competency threshold to redesign and improve its own architecture recursively? That's where the quadratic and exponential curves come from, morons.
It'll never happen, mark my words, screencap this you brain-damaged fool.
It has no intentionality, no consciousness, no qualia. It might (big doubt) get better and better at simulating (faking) some things that abound on github but that's about it.
Chances are it'll never be good enough to even compete with an Indian immigrant.
You'll move on soon enough and put it in the same drawer next to OCR and will stop calling it AI. You'll move the AI over the next hot thing and you'll further redefine words and forget proper senses in order to sustain hype. IT'S a cult.
It's awful considering these aren't real world problems they are github problems and devin is trained on the entirety of github which explicitly contains the answers.
> Random ass benchmark > GPT-4 doing worse than claude 2 (not opus or sonnet, but 2)
Literally a meme tier marketing scheme. Didn't even include Claude Opus, cause it probably beat them even on this meme benchmark
You realize we have so many human software issues that never get resolved for decades on some software right? File managers on linux cant display pictures and havent been for decades.
>You realize we have so many human software issues that never get resolved for decades on some software right? File managers on linux cant display pictures and havent been for decades.
Why doesn't anyone fix that?
Why doesn't firefox display large thumbnails in the upload window?
Why is the default firefox (snap version) so fricking buggy on ubuntu?
why doesnt anyone fix it
>Okay Devin, fix the screen tearing bug on X11, I have have three monitors running on two different refresh rates, the tearing only appears after I wake my computer from sleep state.
It doesnt matter anymore, compared to AI all human SWE will look like diversity hires
2 months ago
Anonymous
I don't think many white people can think of themselves in the same group of "diversity hires". Even when comparing machines to humans.
These things always happen quicker than people think too. My prediction is that in 5-10 years only the top 1% of developers will have jobs in tech, which wont even be coding jobs anyway because the computers will always code faster. 99% of us will be doing something else.
you already pretty much can't get away with just being a simple codemonkey as a white guy, H1Jeets already oversaturated that path long before OpenAI was even relevant. our only path forward is either to roll your own product as an indie or be the architecture guy that translates non-technical customer asks into granular technical requirements
that or make it on shitcoins
Am I the only person alive that is excited for this kind of thing solely because it'll let me get shitloads of personal projects done without burning myself out? Being able to crank out scripts for Tabletop Simulator on demand with minimal effort for anything I want and starting up a game with my old group would be a dream come true.
>the cons
Who cares?
It's exciting new tech with lots of potential regardless of whether it replaces people or not, and it's just as stupid to doompost about SWEs getting replaced as it is to shit on something this cool just because humans can do better.
For me it was pretty evident the moment chatgpt shit out half decent code. I knew it was only a matter of time till it evolves enough and replaces tech jobs, But a part of me still hoped I would have more time
2 months ago
Anonymous
For me it was just obvious from a social perspective, managers and higher ups hate the power that engineers hold in companies and even the theoretical idea that those engineers can be replaced will be endlessly paraded around as soon as it becomes even slightly viable.
2 months ago
Anonymous
These AI models can ONLY SOLVE that which is already available on Stack Overflow. It just saves you the time of copy-and-pasting code snippets. YOU CANNOT USE AI WITHOUT OVERSIGHT. PERIOD.
2 months ago
Anonymous
>These AI models can ONLY SOLVE that which is already available on Stack Overflow. It just saves you the time of copy-and-pasting code snippets. YOU CANNOT USE AI WITHOUT OVERSIGHT. PERIOD.
For now
2 months ago
Anonymous
>2 more years
We're at a dead end for transformers. Until you have an AI that can learn on-the-fly from examples, it's over.
2 months ago
Anonymous
>AI that can learn on-the-fly from examples
you mean at inference time? that's a matter of lowering the cost
2 months ago
Anonymous
moron Transformers cannot "learn" at inference, period. It can only LARP, it cannot learn a new novel concept during inference. Anything that lacks comprehensive text in the training data the AI simply cannot do.
2 months ago
Anonymous
>Anything that lacks comprehensive text in the training data the AI simply cannot do.
This is false. You can provide it new information that may be used if it stays within the context window.
2 months ago
Anonymous
LMAO
Only because you're too stupid to tell the difference between LARP and comprehension. Try to write a story with it with a novel new character and watch as it devolves into tropes.
2 months ago
Anonymous
There are lots of tasks that don't require novel ideas. You said it can't learn a new concept after training is complete. It can. Just not the way humans do ofc.
2 months ago
Anonymous
Most tasks require novel ideas. That's why you're doing them. Most people aren't copy and pasting code snippets all day, at a certain point you are going to be making something that has never been done before and not doing what has been written in code tutorials. ANYTHING off the beaten path the AI fails at and that includes writing stories with NEW characters. Honestly all this just proves that most of you people are low IQ monkeys.
2 months ago
Anonymous
I hate Black folk sir!
2 months ago
Anonymous
pretty much every special snowflake in any organisation wants things done their own way. pretty much all enterprise software is novel.
2 months ago
Anonymous
It doesn't have to really comprehend anything, it just has to produce code that's like 80% correct and I can fix the weird shit it outputs manually. Massively improves my productivity as dev.
2 months ago
Anonymous
You know what's fricked up? This fricking moronic gorilla Black person probably makes twice what I do. Holy shit. I need to start lying and being a delusional piece of shit.
2 months ago
Anonymous
I was being charitable, since that's simply not realistic for the AI paradigm right now. The vast majority of issues out there are remixes and don't require any novel concepts, guidance during should suffice for those.
2 months ago
Anonymous
Wrong, novel includes a business use case that can be an amalgamation of existing code ideas. Let's say you wanted to make a Flutter-based video game version of Pong, the AI would fail.
2 months ago
Anonymous
For those instances I do believe scale is all you need even if it's not worth it.
2 months ago
Anonymous
You're 1000% correct. Anyone who disagrees doesn't know what they're talking about
2 months ago
Anonymous
It does posses emergent abilities though
2 months ago
Anonymous
Hahahahahahaha absolutely subcontinental
2 months ago
Anonymous
go ahead
2 months ago
Anonymous
>We're at a dead end for transformers.
In one month I have seen the average model context size go from 16k-200k to 100k-1mil. Pick a better cope.
2 months ago
Anonymous
That isn't true. I use AI to write code for me all the time, and by that I mean ChatGPT specifically. That being said, and this is key:
I have to fix it up. It gets me like 80% of the way there and then I have to fix it. My skill as a programmer is still pertinent, at least for now. A non-techie wouldn't be able to get that far with it.
Honestly I'm not even sure most programmers would know how to prompt ChatGPT to get the 80%. I just want it to write all the boilerplate shit for me and let me fix the like one critically incorrect line of code.
It makes it so that I'm basically GPT-4's code reviewer and fixer. And it's wonderful for that.
2 months ago
Anonymous
I think it's the Achilles heel of AI. It doesn't have the ability to create something unique. Only derivatives of people's works. This is not limited to programming btw. But idk. I am just some guy fricking with it in stable diffusion.
sensible take but obviously difficult for most to stay rational and objective about this when their livelihood and comfy WFH job with high slaries are dependant on this technology not developing to a level that makes it possible to replace SWE en mass in the next 10 years.
I mean it applies to me too. I just started working in the field and am learning so much having lots of fun diving deeper into the IT Rabbithole. It kind of feels like I am too late to the party. In the end you are right, though. Still not to say how it pans out and what will happen. Lets hope it stays a tool for Programmers instead of making the job history
I can't fricking wait for machines to implement our software ideas with near perfect optimization. Frick human software, 99% of it is shit and there aren't enough of the best 1% to carry it all. The software future might be better than feared.
>implement our software ideas with near perfect optimization
that's not how stochastic models work, it reproduces the median bloated dogshit solution for any given issue
I'm talking future prospects, even if current transformer models aren't the solution, something better will be discovered. Maybe Lecun's JEPA architecture, maybe something different entirely.
I don't know anyone named Devin that wasn't a complete fricking moron.
What's the deal with these names anyway? Devin, Dillon, Ayden how did these gayboy sounding names get popular?
The same psychopathic autists (mostly neets) who were happy about that are the same ones happy about this. I don't know anyone who actually works in software that was happy about dalle except the most soįbrained manchildren, and, again, autists.
AI isn't close to replacing anyone's jobs and it's morons who are so delusional in thinking that the invention of the shovel killed the hand digger's job. At best AI kills some monkey work, but anyone who has done programming knows that 1) monkey work is soul-sucking and 2) there's usually more work than available labor.
I'm not even worried about that, I just fricking despise the attitude that some little queers have towards it. For some reason they remind me of pedophiles.
Also its definitely replaced some copywriters jobs
Yeah good riddance to those copywriters because they were writing low quality SEO sludge. I know this because I worked at a hundred million dollar company that hired such SEO writers who just churned out 5000 word barely researched articles.
>Also its definitely replaced some copywriters jobs
copywriting doesn't need to be "correct". it's like art. AI is doing well at making random shit, because it is just random.
>shovel killed the hand digger's job.
there's no human invention comparable to AI. The shovel couldn't stand on it's own and dig while creating copies of itself in at the same time digging faster and more efficiently.
1) AI can dig a whole by itself maybe 13% of the time, oh and only if there is already a hole there
2) AI does not create copies of itself but nice fantasy that the AI that requires 128 GB of VRAM to run can replicate itself.
Devin can maybe write code accurately 13% of time (probably a generous estimate). Merely reporting that facts of this thread. I'm not the one believing the AI is taking er jerbs
Anyways, post body, post proof you're not a NEET.
2 months ago
Anonymous
I'm guessing you haven't heard of Claude 3. Devin is just a one of the many red flags over the past year.
2 months ago
Anonymous
Please come back within 10 minutes with an entirely novel 10,000 line codebase of a project of your choosing using Claude 3.
this shit and this thread makes me think that the AI shit will win once companies start telling >us to build systems that consider AIs. for example, OS's will have to have standardized error outputs for issues, so that AIs will be able to get around them automatically, and those error messages will need to be "translated" by AI's to humans if and only if needed
So many developers in here pretending they are not gonna be affected by this when 90% of developers are coding deadass dumb business applications that are not much more than CRUD
Here's the sweetest irony of deep learning models and their shills >Have zero artistic talent >Also have zero brains for actual STEM >Larp as STEM genius anyway because you "aced" the JS React course you found through your favorite e-celeb's youtube advert >Overhype AI as something that will destroy all those pesky artists that get more pussy than you >Hype deflates like a wet sock as human eye adapts to recognize slop in less than a month >Start overhyping AI as something that will destroy those pesky leet coders that make more money than you >It fails spectacularly at any higher level of system design higher than writing a bogosort and at any language harder than Python >It does achieve success with boilerplate shitter languages like JS >TFW your frontend monkey job is now obsolete while the artsy UXers use AI to automatically implement their mock-ups
Yeah, it's all hype, and it will not get better. >Claude 3 Opus just reinvented this quantum algorithm from scratch in just 2 prompts. The paper is not on the internet yet.
https://twitter.com/GillVerd/status/1764901418664882327
The humiliation ritual is that she codes for you but the code never works and you spend more time fixing her broken code than you otherwise would have just writing it yourself.
Are you guys really afraid that this AI will replace you? There's no way in hell that establish companies would willingly let an AI that is controlled by another company to read their whole codebase. That literally would be suicide.
https://en.wikipedia.org/wiki/Homomorphic_encryption
Homomorphic encryption is a form of encryption that allows computations to be performed on encrypted data without first having to decrypt it. The resulting computations are left in an encrypted form which, when decrypted, result in an output that is identical to that produced had the operations been performed on the unencrypted data. Homomorphic encryption can be used for privacy-preserving outsourced storage and computation. This allows data to be encrypted and out-sourced to commercial cloud environments for processing, all while encrypted.
The AI will still know the context of those encrypted data. The same AI that is controlled by 1 company. No way in hell that banks or any other conpany that prioritize safety and secrecy will use the AI. Unless they can get an AI that they can run on their own server, it's not happening.
that's like saying >there's no way companies will use cloud hosting and give the provider access to their industrial secrets, that would be suicide
the answer is that they did and they will do the same with ai, under the exact same contractual provisions
Coding ai won't fully replace devs, but it will make each a junior dev 10x more productive than a great senior dev, while great senior dev will lose a lot of advantages. The result is obvious: lower headcount and lower salaries.
>Coding ai won't fully replace devs, but it will make each a junior dev 10x more productive than a great senior dev, while great senior dev will lose a lot of advantages. The result is obvious: lower headcount and lower salaries.
That is how it will start till it evolves to replace all tech jobs
You have that backwards. It will make the senior dev way more productive, as he can now just sit there effectively code reviewing what the AI outputs, while the code monkey junior dev will be out of a job, as he is new at this and thus will be making the exact kind of moron mistakes the AI does.
Let me guess - it learned how to automatically classify and close Indian spam PRs to the express js issue tracker? Still, I suppose that's useful. And a welcome change.
>2 more weeks until artists, translators and voice actors get their jobs back
Have they really not? I mean is there data on what % of them have been 'replaced'? Even if they are not 'replaced' is there quantitative data and statistics about how much it has done for that?
>being this high on hopium
For all the years of hype and thousands of slop threads, imagegen models have failed to make a dent in the massive furry and weeb art trade over at Patreon.
The market has spoken: the genuine article gets the money, the copy-pasted amalgamation doesn't even get any attention.
>Devin is a tireless, skilled teammate, equally ready to build alongside you or independently complete tasks for you to review. >for you to "review"
lmao it's actually a digital pajeet. >fix the cut-rate jeetcode whiteboy!
all the low tier midwits LARPing as geniuses on BOT are in for a rude awakening, you don't even know about AlphaCode 2, and that was a year and a half ago
programmers and eventually all of STEM barring niche engineering, will be 100% replaceable
I came here to shit on you morons for gloating over artists getting their work stolen months ago and I am back again. For people supposedly so smart a lot of you are a bit slow on the uptake that the people who would gladly replace a real human piece of art for machine slop would be more than excited to do so again for a position paid a real salary. They even preemptively washed the job markets forcing you out of good jobs and back into jr. positions just so they would be in a better position to replace your jobs with automation if possible.
This isn’t the end of your careers, it is the slow trudging death of the job market’s bubble. They’re going to keep trying to replace you, more than an artist or a writer. It only makes sense to. You cost much more. Programming will never die though, it’s just that most of you will struggle to get jobs as code janitors in the machine slop era. You can cope over how it’s not here yet but the money someone has the chance to make when it’s “here” is astronomical. Everyone has the profit incentive of the century to make it be here. What are you against the tide of change lead by the predictions of algorithm trading robots? They’ve deduced that we need more robots, the market demands more robots. And you are going to write the code for the bot that replaces you.
We are entering a brave new world. I hope you enjoy it. It’s the world you made.
>It’s the world you made.
geg
codemonkeys cant into ai
let alone inventing something new.
and they are the ones getting replaced.
just like ink monkeys.
genuinely intelligent people already saw the writing on the wall years ago, this place is full of coping fricking idiots
this is just what they present to the public, imagine the shit the alphabet agencies are forcing google and co. to keep under wraps
no work is safe, and that's a good thing
this system was already failing and we need something new
>this system was already failing and we need something new
like solidifying what has happened until now?
because ai will only consolidate the current management...
genuinely intelligent people already saw the writing on the wall years ago, this place is full of coping fricking idiots
this is just what they present to the public, imagine the shit the alphabet agencies are forcing google and co. to keep under wraps
no work is safe, and that's a good thing
this system was already failing and we need something new
AI is just a symptom of a much larger problem that gets sidelined by all the flashy topics: automation. What's happening now isn't anything new. The assembly line replaced a huge amount of jobs, the camera replaced a huge amount of jobs, the typewriter replaced jobs, etc. The only thing that's really new now is the pace that it's happening. The faster everything gets automated, the faster people will have nowhere to turn.
To be honest, I feel like the only way to fix this is with some sort of government intervention that protects the little guy from his robot overlords. How else can you win against automation? And to be clear, I don't think any jobs will fully disappear. Even commerical planes, which basically fly themselves (sans anything made by Boeing in the last decade, kek), still have pilots. Rather, the availability of jobs decreases. Ultimately, I hope we reach a Wall-e situation sometimes in the future. Then everyone can be a permanent NEET. But I imagine it's the getting to that point that's the tough part.
>I feel like the only way to fix this is with some sort of government intervention that protects the little guy from his robot overlords.
until you meet the "little guy" and realize making a good portion of humanity die of hunger, is actually a good thing for the human family
we btfod mother nature but turns out natural selection is what prevented us from going full degen
You forget that countless other jobs unlock because people can just connect the dots at a higher level of thought and make things work or try for a profit, they will always try new things for a profit and it ought to stick, our system incentivizes the invention of fake desires, anything can be hyped and sold, even this meme toy LLM and giant server farm fever dream generators.
>AIAIAIAIAIAIAIAIIAIAIAIAIAIIAIAIAIAIAIAIAI >SWE WILL BE REPLACED IN 2 MORE WEEKS I PROMISE THIS TIME!!!!!! >THIS IS THE END OF CODING >END OF JOBS???? HUMANS REPLACED BY AI IN 2025
dude shut the frick up nobody will ever get fired because of this memey shit. I'm so sick and fricking tired of these clickbaits on youtube gatekeeping motherfrickers. Only a total mouthbreather will take it seriously. Unless we invent matrix-tier machines (which will never happen because we nuke each other long before that) nobody will lose any job
layoffs have already been happening and Devin was already used in paid upwork jobs
you are running out of copes and getting exposed for the midwit you are every single quarter now
Most software jobs can already be done by a high schooler, I doubt they just get rid of the whole control/incentive system that is employment and the risk of being unemployed and homeless.
I've had a prof, from a communications course oddly enough, state that the smart ones will know how to leverage it. Shoola are essentially accepting it as reality.
100% this means lower salaries already. The golden times where doing a bit of webshit landed you comfy working conditions and a above-average salary is OVER 100%
In the next 2 years we will see a big drop in salaries paid guaranteed.
Worst case in the next 10 years most of devs can be replaced and only few remain that manage architecture but thats uncertain
Who's gonna manage an empty office? What projects will even be needed at that point? What once would have taken many departments at a large business can eventually be done by high schoolers.
This shit looks pretty nuts to me if you read this thread of what it did with a limited prompt.
https://twitter.com/itsandrewgao/status/1767893329394413688
Unless it's perfect, it's as useful as the person proooompting. Time and again AI is just provwn to be stackoverflow with less steps. Into the trash it goes
Between Devin, figure, and deepminds new game playing AI (can play entirely unseen games), it's clear what's on the (very short term) horizon. People aren't ready.
>xitter blue
these are the same people who were spouting nft supremecy a few years ago. you giving them attention is exactly what they do this for. frick off back to xitter.
Just so we are clear on the level this shit is improving.
AI art was supposed to fix the hands in like 2 months after SD 1.5 released, according to the same NFT supremacists and it still cannot consistently produce anatomically correct hands or legs.
LLM coding models were barely able to create anything worthwhile with simple instructions, referenced outdated libraries and hallucinated non existent NPC packages by the time GPT 3.5 rolled out.
All of these problems are fixed, and now we have self-correction of code which many thought was not possible and directly interacting with the CLI.
VC money is pouring on coding tools like shit on a New Delhi Street.
It ain't looking good to be honest.
It is in the interest of xitter blue subscribers to increase impressions. they will naturally post clickbait trash like
https://i.imgur.com/WnQcThJ.png
This shit looks pretty nuts to me if you read this thread of what it did with a limited prompt.
https://twitter.com/itsandrewgao/status/1767893329394413688
so engineers will interact with the post and drive engagement.
Literally all of this is done from within a vacuum and everyone is taking it at face value. Literally falling for xitter clickbait.
The other part to note is that this is the worst that these AI models will ever be. They will continue to improve (and likely will improve rapidly based on how quickly progress has been in the past 2 years).
Not saying it's a profound statement, but I think it deserves to be said given all the naysayers who are shitting on the current technology.
> their product looks to me like nothing but a badly edited demo video of someone trying to build a product off AutoGPT / RAG and Chain-of-Thought techniques. There's nothing here models haven't been doing already. They don't appear to even have their own model. Their website is such absolute trash it's tempting to wonder if they had Devin build it.
Damn, some random uninformed moron on reddit got some upvotes, better pack it up
2 months ago
Anonymous
>reddit is wrong because they say it’s bad >twitter is right because they say it’s good
how does the cool aid taste?
2 months ago
Anonymous
reddit is on some social justice crusade against AI because it's unethical
2 months ago
Anonymous
The guy on reddit didn't make any legitimate criticisms with any firsthand knowledge, why would I value that over looking at user videos of what the software actually does? All he said was that it looks like a GPT wrapper, and they had a bad website. Sure the website isn't great, but the company Cognition has only existed for little more than 2 months, and it looks like they spent that time building the product.
2 months ago
Anonymous
it's pretty clear
https://i.imgur.com/BHgbJd1.jpg
It didn't gloss over the primer part though. In his last two tweets, Devin identified the top 10 primer candidates and scored them.
https://twitter.com/itsandrewgao/status/1767896715405791357
>all of these problems are fixed
I'll believe it when I see it in action, cherrypicked examples a la SORA ain't worth shit. >VC money is pouring
That does not mean what you think it means.
This is impressive but I think people might be over rating it here, it seemed to just gloss over the primer part and reduce his prompt to "most recent flu data" in a very search engine like manner. Even Andrew has to kind of gloss over this by going "it implicitly determined I would need to know about the most recent flu strain to know about the primer" no dude you literally have flu strain in the prompt.
It didn't gloss over the primer part though. In his last two tweets, Devin identified the top 10 primer candidates and scored them.
https://twitter.com/itsandrewgao/status/1767896715405791357
> their product looks to me like nothing but a badly edited demo video of someone trying to build a product off AutoGPT / RAG and Chain-of-Thought techniques. There's nothing here models haven't been doing already. They don't appear to even have their own model. Their website is such absolute trash it's tempting to wonder if they had Devin build it.
This one has better discussion even tho it got jannied
https://www.reddit.com/r/cscareerquestions/comments/1bd12gc/relevant_news_cognition_labs_today_were_excited/
Tldr: Devin is prob a complete scam getting shilled here but llm are still scary (everyone already knew that)
Either a bunch of genius level coders for toghether to produce a grift in which case you should be worried or a bunch of genius level coders for together to produce a huge advancement in ai software development in which case you should also be worried.
can i have sex with it?
if not its a useless technology
it would only replace useless prostitutes who are there to look pretty but i cant have sex with it
>spent all the skill points on art and programming >both are being threatened by generative AI
I don't know what to do anymore, man.
I recently picked up gardening and baking. At least I might be able to feed myself after the AI apocalypse.
if youre better than ~90% of human programmers you wont be replaced
once most programmers are replaced theres less incentive to further develop the systems to replace the remaining ~10%
the remaining programmers can then use code generation AI to be more productive
youre not directly competing with AI, in fact it will make you more productive and higher value
Art seems to be doing alright for the most part, jobs are stable and AI art is pretty niche.
If you enjoy art, just keep going, focus on developing a unique style, something catchy and technically impressive.
If you prefer compsci, then just change specialization, cybersecurity is becoming very interesting, and the IMF forecast for US jobs say that ML and Data Science related jobs are getting a massive boost in demand.
will bot ever recover?
Nope. AI is already better at hello world programs than 99.99% of BOT.
Fizzbuzz too, damn.
fa/g//g/ots how are we gonna recover from this?!
>fizzbuzz
>nested elseif
I'm just a scripter and I'd still cut that interview short
what's the issue with that solution?
You can do the same thing with conditionals in 1 line.
...you mean inline conditionals? that's a better solution? I hope you don't actually interview anybody.
Can you explain why efficient recursion is bad before resorting to ad homs, or is it safe to ignore you at this point?
Not that anon, but you do know that recurssion has added cost for no benefit in this case? Compact code isn’t better code. Performant code that is readable and easily modifiable is good code.
>opinion
i can't do much with this
And neither can anybody do much with your code.
hello sir, welcome to brampton
>now using recursion instead of the loop
it's bad because a loop is simpler and recursion creates a dependency that transcends the syntax (loop actions are contained inside the indented block, but recursion goes back to the function definition).
It makes it harder to write, harder to reason about, harder to change in the future.
The version with if...elif is perfection.
You will take it as an hominem, but I used to be like you when I was 15 and writing 'smart' solutions in haskell, lisp and all that. Overengineering is the main problem with junior devs because they make challenges for themselves out of simple problems. With time you will learn that most times, a 'dumber', more verbose is better because it's incredibly easier to debug it and change something.
Recusion only has a reason to exist when it's an inherent property of the problem. Tail recursion works = use a loop. Problem has a nested repeating structure that can't be done as a loop, without emulating recursion at least (eg. a queue) = recursion probably the best solution.
Although there are valid reasons to emulate recursion sometimes, because default stack limit may be too low and changing it may be impractical. Depends on the language.
>simple recursive statement
>transcending syntax
>can't parse a basic b***h fizzbuzz in 1 line
>needs dozens of them and a loop to do the same
I'm sure the women's code bootcamps love you. As far as I'm concerned, you can leave my office now.
Anyone with actual coding experience just rolls your eyes at you. You will roll your eyes at yourself now in a few years time.
Peak midwit.
You are so smooth brained it hurts. You can leave my office now, anon.
Over engineering is the worst blight on software. Devs who over engineer don’t understand how much damage they are causing when they spend more time than necessary to solve a task in a more complex way than is needed on the off chance that they were able to predict the future accurately. Which they never can. So in best case you get a nuke to blow up a treestump. In the best worst case you have to rewrite everything when new requirements come in meaning that their overengineered solution is actively detrimental. And in the worst worst case, also known as the typical case, the overengineered solution stays and everyone has to engineer around it to solve the actual requirements.
Yes. The rule is simple, code for the current problem
>it may need X and Y in the future
no, frick off
>it's better to make it more flexible by using an abstraction
no
sadly, this is something that can only be learned by suffering from mistakes
Wise men at least learn by suffering from the mistakes of others.
In my case it was a combination of semi-justified arrogance
>be genuinely smarter than almost everyone (3+ std in iq tests, and real life situations requiring intelligence)
>treat normal problems as a puzzle
>realize only much later when I need to change something I hurt myself once future requirements go beyond the structure I designed before
being the smartest person in a room is bad for everyone involved
Its imperative that your manager not be able to think he understands the work you're doing or your job is in peril
Funny thing is that I’m the manager. But because I’m also a dev at heart I know that I know all the shit they are doing and the best solutions, but also to stay the frick out because I’m not them, I’m using old data and knowledge and I’m not getting paid to code.
I also know how all dynamics changed when I became a manager. People I could talk with before now get the deer in headlights look when I ask innocent questions out of pure curiosity.
Its just the unfortunate reality of work that alongside all the interpersonal functions a manager provides like guidance and perspective they are also the slave driver who decides if the wagies are doing enough work to stay employed so when they ask work type questions people are always defensive. I'm sure all the layoffs aren't helping that dynamic either.
Ask me how I know you're employed
Ask me how I know you're unemployed
%:include <iostream>
int main() <%
auto bitand o = std::cout;
<::>(auto f) <% f(f, 1); %>(
<::>(auto f, int n) -> void <%
n <= 100 and (
((n % 3 || not(o << "Fizz")) bitand (n % 5 || not(o << "Buzz"*~~ and
o << n,
o << 'n',
f(f, n + 1), 0);
%>);
%>
Are you happy now you piece of shit?
copilot is in fact not filtered by OOP
can you say the same about yourself, BOT?
interface FizzBuzzStrategy {
boolean matches(int number);
String getOutput();
}
class Fizz implements FizzBuzzStrategy {
public boolean matches(int number) {
return number % 3 == 0;
}
public String getOutput() {
return "Fizz";
}
}
class Buzz implements FizzBuzzStrategy {
public boolean matches(int number) {
return number % 5 == 0;
}
public String getOutput() {
return "Buzz";
}
}
class Number implements FizzBuzzStrategy {
private int number;
public Number(int number) {
this.number = number;
}
public boolean matches(int number) {
return true;
}
public String getOutput() {
return Integer.toString(number);
}
}
public class FizzBuzzGame {
private int start;
private int end;
private List<FizzBuzzStrategy> strategies;
public FizzBuzzGame(int start, int end) {
this.start = start;
this.end = end;
this.strategies = Arrays.asList(new Fizz(), new Buzz());
}
public void play() {
for (int i = start; i <= end; i++) {
String result = "";
for (FizzBuzzStrategy strategy : strategies) {
if (strategy.matches(i)) {
result += strategy.getOutput();
}
}
if (result.isEmpty()) {
result = new Number(i).getOutput();
}
System.out.println(result);
}
}
public static void main(String[] args) {
FizzBuzzGame game = new FizzBuzzGame(1, 100);
game.play();
}
}
are you telling us copilot can write ENTERPRISE™ quality code?
>meme dependency injection that fricks up the result if you register the Buzz strategy before the Fizz one
>handling the FizzBuzz with concatenation after all that OOP fanfare
>Number class implements FizzBuzzStrategy to handle the case is which no other strategy matches, but then this case is managed separately making this inheritance unnecessary
Pajeet cargo cult tier.
>cargo cult
Oh my friggin reddit! You grok [0] what "cargo cult" means?! You're so fr*kin cool and smart.
[0] "Grok", a verb coined by Robert A. Heinlein in his 1961 science-fiction novel Stranger in a Strange Land to describe a form of understanding
ty for the book recommendation kind stranger!
>immediate seething reply
It's a funny toy regurgitating silly github code, why aren't you having a laught at the silly code?
The corpocuck style code is silly and stupid but so is your manner of speech. I can think of few terms more quintessentially reddit/hackernews/midwit than "cargo cult." Maybe duck typing and grok.
>I don't like a word you used therefore you're reddit
Tongue my anus, Black person.
I feel like "midwit" is doing a lot of work in that sentence. Do better.
Are you on your period?
There's nothing more reddit than talking about product X and then dropping a "enter product X" for no fricking reason other than being a redditarded reddit Black person.
because what if the silly code isn't silly and i'm just dumb
>>meme dependency injection that fricks up the result if you register the Buzz strategy before the Fizz one
first, there's no dependency injection - it's a hard-coded list in the class where it's used (though I did ask for one in a later question)
second, if you're going to be absurd like that, modification to any implementation will frick up results if you change the order of things unless you also implement explicit ordering/priority
the FizzBuzz with concatenation after all that OOP fanfare
you can ask to refactor that further into OOP, works quite well (did it in a subsequent question)
class implements FizzBuzzStrategy to handle the case is which no other strategy matches, but then this case is managed separately making this inheritance unnecessary
it's actually because I asked for a few back-and-forth changes
copilot's smart enough to do minimal changes to the code rather than completely bulldoze and rewrite everything differently at the smallest change request
previous implementation didn't use concatenation, had strategies for FizzBuzz, Fizz, Buzz and Number, and just checked for the first matching strategy to get the output
yes
no
>% 3 and % 5
>not %15
Kek. Looks like I'm still employed
I really don't get the idea of this ai stuff taking over software engineering jobs. software engineering requires you to use your reasoning skills and problem solving skills to make software. All these "AI" do is just take testing data and use that to predict what word comes after the next. even in an example like this it just takes the training data and copies it. it even left in the example comment on the fortran section and it even says "welcome to w3 schools" proving that it just parrots what it's given. these are impressive tools but if it keeps operating on the same text prediction model that all other ai use, it's just another gimmicky ai product.
what's with the w3schools?
don't take that out, the code needs it
just ignore that. ai doesn't simply scrape content from the web and regurgitate it like some kind of search engine, it's actually learning exactly the same way you and i do. your brain is nothing but a primitive neural network without any adaptable structural hierarchy or recursive phasing. ai is amazing and it's going to take your job chud.
>adaptable structural hierarchy or recursive
Why can't someone just figure out a technique that copies everything a brain can do? It seems like it would be simple
Woah why didn't we think about this before? I'm gonna go try this out in the lab! You're so smart!
I meant, why is it hard? Why doesn't that work or why isn't there a way to do that?
You're so right, it should be easy! We just weren't trying hard enough! It's so obvious!
you're asking why is (human) consciousness hard to recreate, which amounts to why is it hard to create life. moron
I mean, sure. Why would it be hard to recreate (this part of) life?
LISP: "Welcome to W3Schools" hmmmm
>printf
>not puts
so ai writes jeetcode?
At least they know ciobol.
>using namespace std
Into the trash it goes.
I solve 100% of the bugs.
yeah no
>0.52 to 13.86 in less than 4 years
>WE'RE GOOD BRO
Nobody's saying it isn't coming eventually, but this is still not it
100% in 1 year.
>YEAH BRO THIBGS JUST GET EXPONENTIALLY BETTER NON STOP FOREVER
With how many billions the people of coin have poured into the devil that is AI, this is a laughable improvement, if you can even call this "an improvement" at all.
>we've gone from "erm... it'll never improve" to downplaying improvements
lolz
Is it really downplaying improvements when the shit boomed like six years ago and it can't even replace my local Rajeesh coding for rent money with ten of his benchods yet?
>boomed like six years ago
More like a year ago. Before that it was only the AI from "if else" memes.
before that AI was just called Machine Learning
>it'll replace me in 2 years instead of 1.5 years, check that!
That benchmark is or AI doing SWE tasks unassisted. With human assistance it will be able to do the tasks near 100%. This would mean fewer job opportunities with SWE roles
No, because sadly every time we've invented a way to somehow need less SWEs to do things, we just end up needing more SWEs because less viable projects become more viable.
Previously less viable projects increasing viability would still require way fewer human SWEs
But the number of projects would increase.
This is what's happened every time we've massively increased SWE productivity.
>But the number of projects would increase.
Sure but the number of human SWE wont
actually, it will and that is the part which most recognize is the worst.
the iq barrier for entering the SWE domain is becoming lower and lower.
web devs are the Black folk of SWE world, but soon, SWE will become the Black person of engineering, because it will be flooded with single digit "developers", like the web was flooded with "hTmL/CsS engineers".
go to youtube and watch completely tech agnostic people trying to spin up a "10k/mmr saas project".
every single step is a prompt input.
these are subhumans we are talking about..
there are normal, smart engineers, then there are non-individuated NPCs, who usually work at frontend.. but this class now is even below that, they barely qualify as human, only in physical appearance.
SWE is doomed.
question is, how fast will other domains be affected.
this fake AI can be trained to learn how to optimally control those robot hands with which some hospitals perform surgery.
surgery could be completely automated away, potentially.
this is in an ideal future, where the stupid AI does not hallucinate shit and can solve unexpected problems that arise on the fly, due to is emergent "smart" properties.
https://en.wikipedia.org/wiki/Jevons_paradox
Just like how high level languages led to fewer jobs than when programming was done in assembly. More productivity = fewer jobs, because the amount of work to do is constant and making programming quicker doesn't increase the demand for programming.
>Just like how high level languages led to fewer jobs than when programming was done in assembly
When high level languages started becoming mainstream the tech sector itself was niche.
AI is becoming mainstream to an already mainstream tech sector with capacity to cause large job displacement. I'm sure AI will create more jobs in the future but the transition is going to be pretty rough
No because Devin is designed to be human-free. This is why Devin crushes all the competition, because its a mouse competing against fish in swimming contest, absolutely not comparable
>because its a mouse competing against fish in swimming contest, absolutely not comparable
Oh, dear, another bot post
In another 4 years it's gonna be 1447.69
:O
>14% accuracy after blatantly training on the test data
Yeah, we're good.
>>14% accuracy after blatantly training on the test data
By a little known AI startup. OpenAI is gonna up the numbers and other AI players are gonna try to catch them within an year
Training on test data means you're literally just testing memorization, put some effort into your trolling.
It should be illegal to have this homosexual west coast way of writing. Corpogays should die.
>this AI went from super duper trash to super trash
>therefore SWE will go extinct
i could piss on the server and the AI will stop working.
Thin is yea it'll reach a plateau that's impressive BUT for the most basic things that are spammed across github, for everything novel, curveball problem and so on it'll forever be a random solution generator.
Yeah, same thing happened with computer hardware. Oh wait...
yea good point, Moore's Law died around 2015, they already knew by late 1990s how milk the last bit of possible performance not wiht gate size but with its height lol solely with density grind which again it kind of already reached its limit
You do realize compute has been growing exponentially even more than Moore's law
yeah it did lol
look at the 3090 vs the 4090
its just much bigger lol
a sigma function with a high enough plateau is indistinguishable from an exponential on a given scale.
For real. People forget the industrial revolution was built off the back of essentially a single thing: converting stored energy to automatic movement. Sure, we hit the top of the curve eventually, but it took almost a hundred years and utterly transformed humanity.
Yeah but what about a ligma function?
ligma what now?
Ligma balls
who's steve jobs?
How will you midwits cope the moment a model achieves the competency threshold to redesign and improve its own architecture recursively? That's where the quadratic and exponential curves come from, morons.
It'll never happen, mark my words, screencap this you brain-damaged fool.
It has no intentionality, no consciousness, no qualia. It might (big doubt) get better and better at simulating (faking) some things that abound on github but that's about it.
Chances are it'll never be good enough to even compete with an Indian immigrant.
You'll move on soon enough and put it in the same drawer next to OCR and will stop calling it AI. You'll move the AI over the next hot thing and you'll further redefine words and forget proper senses in order to sustain hype. IT'S a cult.
>qualia
KEK
>KEK
Ah I see that you are a body type 1 "female".
https://plato.stanford.edu/entries/qualia/
It has some funny meaning in the transgender community?
We're reaching levels of copium that shouldn't even be possible.
>t. clueless zoomer nocoder
if this happens then we are either all dead/tortured or living a utopia. your job doesn't matter in this case.
How does it fare against copilot?
that seems pretty good tbh. that being said I have no idea what you're even talking about
It's awful considering these aren't real world problems they are github problems and devin is trained on the entirety of github which explicitly contains the answers.
It's the AI solving problems relating to programming with no assistance whatsoever
> Random ass benchmark
> GPT-4 doing worse than claude 2 (not opus or sonnet, but 2)
Literally a meme tier marketing scheme. Didn't even include Claude Opus, cause it probably beat them even on this meme benchmark
It doesn't have to be as good as you or I, it just has to compete with your average pajeet (not a high bar to clear) and we'll all be out of jobs.
Which one of these can I use for free to do my job for me?
Lmao it doesn't need to be better than an engineer, it just needs to be pajeet-tier to replace a shitload of current devs
You realize we have so many human software issues that never get resolved for decades on some software right? File managers on linux cant display pictures and havent been for decades.
>You realize we have so many human software issues that never get resolved for decades on some software right? File managers on linux cant display pictures and havent been for decades.
Why doesn't anyone fix that?
Why doesn't firefox display large thumbnails in the upload window?
Why is the default firefox (snap version) so fricking buggy on ubuntu?
why doesnt anyone fix it
Now we just need AI to fork it, fix it, patch it, build it, ship it, etc.
Its gonna be a self evolving system that will constantly fix whatever issues users click.
Just like Sora, Q*, and every other isolated display of performance that's come out since 2023 with no sign of publicization: Fake and Gay.
As long as it impresses and convinces management...
>Okay Devin, fix the screen tearing bug on X11, I have have three monitors running on two different refresh rates, the tearing only appears after I wake my computer from sleep state.
>ok Devin, please drag and drop this item on wayland
>implying humans can reliably fix that kind of random stuff
It's a 50/50 for both human and AI
that's a lot of cum!
So what now? learn to woodwork?
Learn to sabotage electric substations
>he wasn't a woodworker's boy
ngmi
Trades are the only AI-proof jobs. Learn plumbing, mechanics or roofing. Or get a CDL.
real engineering that's not CS cant be replaced by AI.
the only thing these shitty AIs can replace are diversity hires, but they wont get rid of them due to ESG
>I am safe from AI because I am white
The AI storm is gonna hit you like a freight train you done see coming
ok jamal, dont worry im well aware your and karen's jobs are safe thanks to larry fink.
No, moron, All our jobs are at risk
thats where you're wrong. diversity hires could never do any work to begin with, they were forced to hire them and they will continue to do so.
It doesnt matter anymore, compared to AI all human SWE will look like diversity hires
I don't think many white people can think of themselves in the same group of "diversity hires". Even when comparing machines to humans.
These things always happen quicker than people think too. My prediction is that in 5-10 years only the top 1% of developers will have jobs in tech, which wont even be coding jobs anyway because the computers will always code faster. 99% of us will be doing something else.
you already pretty much can't get away with just being a simple codemonkey as a white guy, H1Jeets already oversaturated that path long before OpenAI was even relevant. our only path forward is either to roll your own product as an indie or be the architecture guy that translates non-technical customer asks into granular technical requirements
that or make it on shitcoins
>Devin
Can we go gayer tho? How about Melvin?
It can't do anything that GPT-4 or Claude couldn't already do albeit with a user in the loop. Enjoy your webdev prototype crap.
>with human assistance
Am I the only person alive that is excited for this kind of thing solely because it'll let me get shitloads of personal projects done without burning myself out? Being able to crank out scripts for Tabletop Simulator on demand with minimal effort for anything I want and starting up a game with my old group would be a dream come true.
The pros are great but the cons are scary
>the cons
Who cares?
It's exciting new tech with lots of potential regardless of whether it replaces people or not, and it's just as stupid to doompost about SWEs getting replaced as it is to shit on something this cool just because humans can do better.
>Who cares?
I care, I barely started earning a decent wage. I dont know how long I will be able to have the lifestyle I want or aspire
Did it not occur to you at some point in all the endless skilling up and changing goal posts that something like this was inevitable?
For me it was pretty evident the moment chatgpt shit out half decent code. I knew it was only a matter of time till it evolves enough and replaces tech jobs, But a part of me still hoped I would have more time
For me it was just obvious from a social perspective, managers and higher ups hate the power that engineers hold in companies and even the theoretical idea that those engineers can be replaced will be endlessly paraded around as soon as it becomes even slightly viable.
These AI models can ONLY SOLVE that which is already available on Stack Overflow. It just saves you the time of copy-and-pasting code snippets. YOU CANNOT USE AI WITHOUT OVERSIGHT. PERIOD.
>These AI models can ONLY SOLVE that which is already available on Stack Overflow. It just saves you the time of copy-and-pasting code snippets. YOU CANNOT USE AI WITHOUT OVERSIGHT. PERIOD.
For now
>2 more years
We're at a dead end for transformers. Until you have an AI that can learn on-the-fly from examples, it's over.
>AI that can learn on-the-fly from examples
you mean at inference time? that's a matter of lowering the cost
moron Transformers cannot "learn" at inference, period. It can only LARP, it cannot learn a new novel concept during inference. Anything that lacks comprehensive text in the training data the AI simply cannot do.
>Anything that lacks comprehensive text in the training data the AI simply cannot do.
This is false. You can provide it new information that may be used if it stays within the context window.
LMAO
Only because you're too stupid to tell the difference between LARP and comprehension. Try to write a story with it with a novel new character and watch as it devolves into tropes.
There are lots of tasks that don't require novel ideas. You said it can't learn a new concept after training is complete. It can. Just not the way humans do ofc.
Most tasks require novel ideas. That's why you're doing them. Most people aren't copy and pasting code snippets all day, at a certain point you are going to be making something that has never been done before and not doing what has been written in code tutorials. ANYTHING off the beaten path the AI fails at and that includes writing stories with NEW characters. Honestly all this just proves that most of you people are low IQ monkeys.
I hate Black folk sir!
pretty much every special snowflake in any organisation wants things done their own way. pretty much all enterprise software is novel.
It doesn't have to really comprehend anything, it just has to produce code that's like 80% correct and I can fix the weird shit it outputs manually. Massively improves my productivity as dev.
You know what's fricked up? This fricking moronic gorilla Black person probably makes twice what I do. Holy shit. I need to start lying and being a delusional piece of shit.
I was being charitable, since that's simply not realistic for the AI paradigm right now. The vast majority of issues out there are remixes and don't require any novel concepts, guidance during should suffice for those.
Wrong, novel includes a business use case that can be an amalgamation of existing code ideas. Let's say you wanted to make a Flutter-based video game version of Pong, the AI would fail.
For those instances I do believe scale is all you need even if it's not worth it.
You're 1000% correct. Anyone who disagrees doesn't know what they're talking about
It does posses emergent abilities though
Hahahahahahaha absolutely subcontinental
go ahead
>We're at a dead end for transformers.
In one month I have seen the average model context size go from 16k-200k to 100k-1mil. Pick a better cope.
That isn't true. I use AI to write code for me all the time, and by that I mean ChatGPT specifically. That being said, and this is key:
I have to fix it up. It gets me like 80% of the way there and then I have to fix it. My skill as a programmer is still pertinent, at least for now. A non-techie wouldn't be able to get that far with it.
Honestly I'm not even sure most programmers would know how to prompt ChatGPT to get the 80%. I just want it to write all the boilerplate shit for me and let me fix the like one critically incorrect line of code.
It makes it so that I'm basically GPT-4's code reviewer and fixer. And it's wonderful for that.
I think it's the Achilles heel of AI. It doesn't have the ability to create something unique. Only derivatives of people's works. This is not limited to programming btw. But idk. I am just some guy fricking with it in stable diffusion.
>Who cares?
>It's exciting new tech with lots of potential regardless
Face the wall
sensible take but obviously difficult for most to stay rational and objective about this when their livelihood and comfy WFH job with high slaries are dependant on this technology not developing to a level that makes it possible to replace SWE en mass in the next 10 years.
I mean it applies to me too. I just started working in the field and am learning so much having lots of fun diving deeper into the IT Rabbithole. It kind of feels like I am too late to the party. In the end you are right, though. Still not to say how it pans out and what will happen. Lets hope it stays a tool for Programmers instead of making the job history
I can't fricking wait for machines to implement our software ideas with near perfect optimization. Frick human software, 99% of it is shit and there aren't enough of the best 1% to carry it all. The software future might be better than feared.
>implement our software ideas with near perfect optimization
that's not how stochastic models work, it reproduces the median bloated dogshit solution for any given issue
>median bloated dogshit solution for any given issue
AKA what we have now
yes, exactly. for industrial scale deployment of "near perfect optimization" we'd need the mystical future AGI and not die
I'm talking future prospects, even if current transformer models aren't the solution, something better will be discovered. Maybe Lecun's JEPA architecture, maybe something different entirely.
There is some university startup that using worm brain architecture for better cognitive approach. I can see that payoff
Lmao it hasn’t even been a year since smug BOT software engineers said this will never happen.
THINK OF HOW GOOD IT'LL BE IN ONE YEAR
I don't know anyone named Devin that wasn't a complete fricking moron.
What's the deal with these names anyway? Devin, Dillon, Ayden how did these gayboy sounding names get popular?
>dev in
you were so happy about "btfoing the artists"
seems like you'll be the first ones out of a job
The same psychopathic autists (mostly neets) who were happy about that are the same ones happy about this. I don't know anyone who actually works in software that was happy about dalle except the most soįbrained manchildren, and, again, autists.
AI isn't close to replacing anyone's jobs and it's morons who are so delusional in thinking that the invention of the shovel killed the hand digger's job. At best AI kills some monkey work, but anyone who has done programming knows that 1) monkey work is soul-sucking and 2) there's usually more work than available labor.
I'm not even worried about that, I just fricking despise the attitude that some little queers have towards it. For some reason they remind me of pedophiles.
Also its definitely replaced some copywriters jobs
Yeah good riddance to those copywriters because they were writing low quality SEO sludge. I know this because I worked at a hundred million dollar company that hired such SEO writers who just churned out 5000 word barely researched articles.
>Also its definitely replaced some copywriters jobs
copywriting doesn't need to be "correct". it's like art. AI is doing well at making random shit, because it is just random.
>shovel killed the hand digger's job.
there's no human invention comparable to AI. The shovel couldn't stand on it's own and dig while creating copies of itself in at the same time digging faster and more efficiently.
1) AI can dig a whole by itself maybe 13% of the time, oh and only if there is already a hole there
2) AI does not create copies of itself but nice fantasy that the AI that requires 128 GB of VRAM to run can replicate itself.
LMAO
take meds
Devin can maybe write code accurately 13% of time (probably a generous estimate). Merely reporting that facts of this thread. I'm not the one believing the AI is taking er jerbs
Anyways, post body, post proof you're not a NEET.
I'm guessing you haven't heard of Claude 3. Devin is just a one of the many red flags over the past year.
Please come back within 10 minutes with an entirely novel 10,000 line codebase of a project of your choosing using Claude 3.
If I can't code I'll go full into AI art.
this shit and this thread makes me think that the AI shit will win once companies start telling >us to build systems that consider AIs. for example, OS's will have to have standardized error outputs for issues, so that AIs will be able to get around them automatically, and those error messages will need to be "translated" by AI's to humans if and only if needed
People in here still denying it. See you guys in 6 months.
I T I S O V E R
So many developers in here pretending they are not gonna be affected by this when 90% of developers are coding deadass dumb business applications that are not much more than CRUD
You need a developer to verify the AI did the code correctly.
Yes, A developer. Not a team. A single one. So one guys gets to keep his job as the AI QA and the other five are out of a job.
Here's the sweetest irony of deep learning models and their shills
>Have zero artistic talent
>Also have zero brains for actual STEM
>Larp as STEM genius anyway because you "aced" the JS React course you found through your favorite e-celeb's youtube advert
>Overhype AI as something that will destroy all those pesky artists that get more pussy than you
>Hype deflates like a wet sock as human eye adapts to recognize slop in less than a month
>Start overhyping AI as something that will destroy those pesky leet coders that make more money than you
>It fails spectacularly at any higher level of system design higher than writing a bogosort and at any language harder than Python
>It does achieve success with boilerplate shitter languages like JS
>TFW your frontend monkey job is now obsolete while the artsy UXers use AI to automatically implement their mock-ups
Yeah, it's all hype, and it will not get better.
>Claude 3 Opus just reinvented this quantum algorithm from scratch in just 2 prompts. The paper is not on the internet yet.
https://twitter.com/GillVerd/status/1764901418664882327
I WOULD ACCEPT TO BE REPLACED IF ALL HR/MANAGERS GET FIRED TOO
Dream come true frfr
Bros which ai should I use to build my trading algo? I’m moronic so I need to be able to feed it pictures to communicate
sell trading algo
don't actually use one
why is devin making that face
is this some sort of humiliation ritual where she codes for me but also bullies and degrades me in each message
The humiliation ritual is that she codes for you but the code never works and you spend more time fixing her broken code than you otherwise would have just writing it yourself.
welp, I'm killing myself, what's BOTs plan?
Can anyone use it yet or are we still on cherrypicked demo videos where I can't see that the AI hallucinated 90% of the time?
A number of people are using it currently based on the threads on twitter, but you need to join a waitlist at the moment if you're interested.
Let me guess. Approved ~~*journalists*~~ right now.
Are you guys really afraid that this AI will replace you? There's no way in hell that establish companies would willingly let an AI that is controlled by another company to read their whole codebase. That literally would be suicide.
https://en.wikipedia.org/wiki/Homomorphic_encryption
Homomorphic encryption is a form of encryption that allows computations to be performed on encrypted data without first having to decrypt it. The resulting computations are left in an encrypted form which, when decrypted, result in an output that is identical to that produced had the operations been performed on the unencrypted data. Homomorphic encryption can be used for privacy-preserving outsourced storage and computation. This allows data to be encrypted and out-sourced to commercial cloud environments for processing, all while encrypted.
The AI will still know the context of those encrypted data. The same AI that is controlled by 1 company. No way in hell that banks or any other conpany that prioritize safety and secrecy will use the AI. Unless they can get an AI that they can run on their own server, it's not happening.
>unless they can get an Ai that they can run on their own server
Those already exist
that's like saying
>there's no way companies will use cloud hosting and give the provider access to their industrial secrets, that would be suicide
the answer is that they did and they will do the same with ai, under the exact same contractual provisions
Coding ai won't fully replace devs, but it will make each a junior dev 10x more productive than a great senior dev, while great senior dev will lose a lot of advantages. The result is obvious: lower headcount and lower salaries.
Are banks using cloud hosting?
>Coding ai won't fully replace devs, but it will make each a junior dev 10x more productive than a great senior dev, while great senior dev will lose a lot of advantages. The result is obvious: lower headcount and lower salaries.
That is how it will start till it evolves to replace all tech jobs
You have that backwards. It will make the senior dev way more productive, as he can now just sit there effectively code reviewing what the AI outputs, while the code monkey junior dev will be out of a job, as he is new at this and thus will be making the exact kind of moron mistakes the AI does.
This thread would be right in place in /ic/ a year ago if you just change the name of the model
ITT: Black folk who never read "No Silver Bullet"
its a scam
https://old.reddit.com/r/cscareerquestions/comments/1bd12gc/relevant_news_cognition_labs_today_were_excited/kujyidr/
Let me guess - it learned how to automatically classify and close Indian spam PRs to the express js issue tracker? Still, I suppose that's useful. And a welcome change.
2 more weeks until doctors, lawyers, mathematicians, and engineers are all replaced by computers
2 more weeks until artists, translators and voice actors get their jobs back
>2 more weeks until artists, translators and voice actors get their jobs back
Have they really not? I mean is there data on what % of them have been 'replaced'? Even if they are not 'replaced' is there quantitative data and statistics about how much it has done for that?
>being this high on hopium
For all the years of hype and thousands of slop threads, imagegen models have failed to make a dent in the massive furry and weeb art trade over at Patreon.
The market has spoken: the genuine article gets the money, the copy-pasted amalgamation doesn't even get any attention.
2 more weeks until they actually lose their jobs, you just wait
>software engineer
cringe
So, another copilot?
>Devin is a tireless, skilled teammate, equally ready to build alongside you or independently complete tasks for you to review.
>for you to "review"
lmao it's actually a digital pajeet.
>fix the cut-rate jeetcode whiteboy!
This thing can train its own AI models apparently. Now might be the time to have your Faraday cage installed
is so why this thing would replace only programmer? this will kill all computer job? if so we are doomed
IT'S FRICKING OVER
all the low tier midwits LARPing as geniuses on BOT are in for a rude awakening, you don't even know about AlphaCode 2, and that was a year and a half ago
programmers and eventually all of STEM barring niche engineering, will be 100% replaceable
I came here to shit on you morons for gloating over artists getting their work stolen months ago and I am back again. For people supposedly so smart a lot of you are a bit slow on the uptake that the people who would gladly replace a real human piece of art for machine slop would be more than excited to do so again for a position paid a real salary. They even preemptively washed the job markets forcing you out of good jobs and back into jr. positions just so they would be in a better position to replace your jobs with automation if possible.
This isn’t the end of your careers, it is the slow trudging death of the job market’s bubble. They’re going to keep trying to replace you, more than an artist or a writer. It only makes sense to. You cost much more. Programming will never die though, it’s just that most of you will struggle to get jobs as code janitors in the machine slop era. You can cope over how it’s not here yet but the money someone has the chance to make when it’s “here” is astronomical. Everyone has the profit incentive of the century to make it be here. What are you against the tide of change lead by the predictions of algorithm trading robots? They’ve deduced that we need more robots, the market demands more robots. And you are going to write the code for the bot that replaces you.
We are entering a brave new world. I hope you enjoy it. It’s the world you made.
>It’s the world you made.
geg
codemonkeys cant into ai
let alone inventing something new.
and they are the ones getting replaced.
just like ink monkeys.
genuinely intelligent people already saw the writing on the wall years ago, this place is full of coping fricking idiots
this is just what they present to the public, imagine the shit the alphabet agencies are forcing google and co. to keep under wraps
no work is safe, and that's a good thing
this system was already failing and we need something new
>this system was already failing and we need something new
like solidifying what has happened until now?
because ai will only consolidate the current management...
AI is just a symptom of a much larger problem that gets sidelined by all the flashy topics: automation. What's happening now isn't anything new. The assembly line replaced a huge amount of jobs, the camera replaced a huge amount of jobs, the typewriter replaced jobs, etc. The only thing that's really new now is the pace that it's happening. The faster everything gets automated, the faster people will have nowhere to turn.
To be honest, I feel like the only way to fix this is with some sort of government intervention that protects the little guy from his robot overlords. How else can you win against automation? And to be clear, I don't think any jobs will fully disappear. Even commerical planes, which basically fly themselves (sans anything made by Boeing in the last decade, kek), still have pilots. Rather, the availability of jobs decreases. Ultimately, I hope we reach a Wall-e situation sometimes in the future. Then everyone can be a permanent NEET. But I imagine it's the getting to that point that's the tough part.
>I feel like the only way to fix this is with some sort of government intervention that protects the little guy from his robot overlords.
until you meet the "little guy" and realize making a good portion of humanity die of hunger, is actually a good thing for the human family
we btfod mother nature but turns out natural selection is what prevented us from going full degen
You forget that countless other jobs unlock because people can just connect the dots at a higher level of thought and make things work or try for a profit, they will always try new things for a profit and it ought to stick, our system incentivizes the invention of fake desires, anything can be hyped and sold, even this meme toy LLM and giant server farm fever dream generators.
>AIAIAIAIAIAIAIAIIAIAIAIAIAIIAIAIAIAIAIAIAI
>SWE WILL BE REPLACED IN 2 MORE WEEKS I PROMISE THIS TIME!!!!!!
>THIS IS THE END OF CODING
>END OF JOBS???? HUMANS REPLACED BY AI IN 2025
dude shut the frick up nobody will ever get fired because of this memey shit. I'm so sick and fricking tired of these clickbaits on youtube gatekeeping motherfrickers. Only a total mouthbreather will take it seriously. Unless we invent matrix-tier machines (which will never happen because we nuke each other long before that) nobody will lose any job
jhonny close BOT and go help fight the arasaka
layoffs have already been happening and Devin was already used in paid upwork jobs
you are running out of copes and getting exposed for the midwit you are every single quarter now
layoffs aren't happening because of AI
you have no idea what you are talking about mate
Most software jobs can already be done by a high schooler, I doubt they just get rid of the whole control/incentive system that is employment and the risk of being unemployed and homeless.
How is the mood at schools? What are the teachers saying? What about the students?
How is everyone coping?
I've had a prof, from a communications course oddly enough, state that the smart ones will know how to leverage it. Shoola are essentially accepting it as reality.
it's over
100% this means lower salaries already. The golden times where doing a bit of webshit landed you comfy working conditions and a above-average salary is OVER 100%
In the next 2 years we will see a big drop in salaries paid guaranteed.
Worst case in the next 10 years most of devs can be replaced and only few remain that manage architecture but thats uncertain
What about the project management?
Who's gonna manage an empty office? What projects will even be needed at that point? What once would have taken many departments at a large business can eventually be done by high schoolers.
ur moronic
freetards gave not only their software but also their ability to develop software to corporations for free.
Imagine being a programmer and knowingly coding something that will replace you
Top tier cuck behavior
This shit looks pretty nuts to me if you read this thread of what it did with a limited prompt.
https://twitter.com/itsandrewgao/status/1767893329394413688
>I fixed it
Unless it's perfect, it's as useful as the person proooompting. Time and again AI is just provwn to be stackoverflow with less steps. Into the trash it goes
It's actually
>it fixed it
It can catch its own mistakes and rewrite the code
>they figured out how to plug together a few api's and get an llm to prooompt itself
wow
if you read through his other tweets you'll see he had to point out mistakes in other places
Reread what he said. The twitter guy didn't fix the error, Devin found the error and fixed it on its own, then re-ran the code.
holy shit it's habbening!
Between Devin, figure, and deepminds new game playing AI (can play entirely unseen games), it's clear what's on the (very short term) horizon. People aren't ready.
>People aren't ready.
they never are.
I'm not sure with this has to do with things that currently, right now exist.
funnily enough that’s exactly what people said about the web in the 90s.
>xitter blue
these are the same people who were spouting nft supremecy a few years ago. you giving them attention is exactly what they do this for. frick off back to xitter.
Just so we are clear on the level this shit is improving.
AI art was supposed to fix the hands in like 2 months after SD 1.5 released, according to the same NFT supremacists and it still cannot consistently produce anatomically correct hands or legs.
LLM coding models were barely able to create anything worthwhile with simple instructions, referenced outdated libraries and hallucinated non existent NPC packages by the time GPT 3.5 rolled out.
All of these problems are fixed, and now we have self-correction of code which many thought was not possible and directly interacting with the CLI.
VC money is pouring on coding tools like shit on a New Delhi Street.
It ain't looking good to be honest.
>It ain't looking good to be honest.
this train is not stopping anytime soon.
It is in the interest of xitter blue subscribers to increase impressions. they will naturally post clickbait trash like
so engineers will interact with the post and drive engagement.
Literally all of this is done from within a vacuum and everyone is taking it at face value. Literally falling for xitter clickbait.
The other part to note is that this is the worst that these AI models will ever be. They will continue to improve (and likely will improve rapidly based on how quickly progress has been in the past 2 years).
>this is the worst that these AI models will ever be
I wish people would stop parroting this as some profound insight. It’s pretty insufferable.
Not saying it's a profound statement, but I think it deserves to be said given all the naysayers who are shitting on the current technology.
Damn, some random uninformed moron on reddit got some upvotes, better pack it up
>reddit is wrong because they say it’s bad
>twitter is right because they say it’s good
how does the cool aid taste?
reddit is on some social justice crusade against AI because it's unethical
The guy on reddit didn't make any legitimate criticisms with any firsthand knowledge, why would I value that over looking at user videos of what the software actually does? All he said was that it looks like a GPT wrapper, and they had a bad website. Sure the website isn't great, but the company Cognition has only existed for little more than 2 months, and it looks like they spent that time building the product.
it's pretty clear
>all of these problems are fixed
I'll believe it when I see it in action, cherrypicked examples a la SORA ain't worth shit.
>VC money is pouring
That does not mean what you think it means.
This is impressive but I think people might be over rating it here, it seemed to just gloss over the primer part and reduce his prompt to "most recent flu data" in a very search engine like manner. Even Andrew has to kind of gloss over this by going "it implicitly determined I would need to know about the most recent flu strain to know about the primer" no dude you literally have flu strain in the prompt.
It didn't gloss over the primer part though. In his last two tweets, Devin identified the top 10 primer candidates and scored them.
https://twitter.com/itsandrewgao/status/1767896715405791357
Yeah that's pretty wild then, twitter just links the one image then asks me to make an account so I only saw the one post
At the same time if you look at the dudes Twitter history he's basically a grifter who got his first job in 2023 Nov so idk
like all AI demos, this is canned / cherrypicked rubbish.
It's OVER for you code monkey Black folk
HAHAHAHAHAHAH
homosexualS!
> their product looks to me like nothing but a badly edited demo video of someone trying to build a product off AutoGPT / RAG and Chain-of-Thought techniques. There's nothing here models haven't been doing already. They don't appear to even have their own model. Their website is such absolute trash it's tempting to wonder if they had Devin build it.
https://www.reddit.com/r/MachineLearning/comments/1bdzesy/thoughts_on_the_latest_ai_software_engineer_devin/
lol
lmao
he knows Devin will replace his ass
This one has better discussion even tho it got jannied
https://www.reddit.com/r/cscareerquestions/comments/1bd12gc/relevant_news_cognition_labs_today_were_excited/
Tldr: Devin is prob a complete scam getting shilled here but llm are still scary (everyone already knew that)
https://twitter.com/mckaywrigley/status/1767985840448516343
why is it so slow tho
They probably don't have the infrastructure in place yet to handle all of the compute required to run these requests quickly.
>14 mins in
>okay i cloned the repo
not that different from the people at work honestly
This Devin model is shit but please understand there will be almost no warning between laughing at shit models and a meeting with HR.
Yup but at the same time that doesn't mean people have to give the air to every scammer who says they are that new career changing model
so much cope about this it's crazy
redditors are so butthurt they're trying to harass the creators
how insecure can you be?
Computers as augmentation vs automation.
You dun goofed.
It's funny how their reactions to this parallel artists' reactions to AI art, which was 1.5 years ago now
Alright I’m not a codegay, is this unironically legit or nothing to worry about
Either a bunch of genius level coders for toghether to produce a grift in which case you should be worried or a bunch of genius level coders for together to produce a huge advancement in ai software development in which case you should also be worried.
can i have sex with it?
if not its a useless technology
it would only replace useless prostitutes who are there to look pretty but i cant have sex with it
>spent all the skill points on art and programming
>both are being threatened by generative AI
I don't know what to do anymore, man.
I recently picked up gardening and baking. At least I might be able to feed myself after the AI apocalypse.
if youre better than ~90% of human programmers you wont be replaced
once most programmers are replaced theres less incentive to further develop the systems to replace the remaining ~10%
the remaining programmers can then use code generation AI to be more productive
youre not directly competing with AI, in fact it will make you more productive and higher value
Art seems to be doing alright for the most part, jobs are stable and AI art is pretty niche.
If you enjoy art, just keep going, focus on developing a unique style, something catchy and technically impressive.
If you prefer compsci, then just change specialization, cybersecurity is becoming very interesting, and the IMF forecast for US jobs say that ML and Data Science related jobs are getting a massive boost in demand.
Garbage in garbage out
So if AI will soon replace pajeet tier workers, does this mean the first world will stop indian immigration?
Devin bros...
Just in time for 4.5-turbo announced later today