GPT-4 Makes Old ChatGPT Look Like a JOKE!
ฝัง
- เผยแพร่เมื่อ 14 มี.ค. 2023
- Check out my courses: dometrain.com
Become a Patreon and get source code access: / nickchapsas
Hello everybody I'm Nick and in this video I will test the brand new GPT-4 (ChatGPT 4) model blindly. I will see if it can answer questions and write code that Junior, Mid, senior or even lead developers would have to answer as part of their day job.
Don't forget to comment, like and subscribe :)
Social Media:
Follow me on GitHub: bit.ly/ChapsasGitHub
Follow me on Twitter: bit.ly/ChapsasTwitter
Connect on LinkedIn: bit.ly/ChapsasLinkedIn
Keep coding merch: keepcoding.shop
#opena #gpt4 #chatgpt
Those saying that it is "just a tool" are right, but it's a tool with knowledge of every single domain and the ability to write good code, very quickly, architect and give scaling advice. That's exactly what developers do and for business people, developers a "just a tool" to get their product out. See where I'm going with this? Denial won't help you adapt to the new world that we will be living in very soon. You will only be replaced if you fail to adapt.
Nick, in the future, what you can also do is tell it to continue from the beginning of method 'x'. This will fix the issue of the code blocks getting thrown off as it will return a statement of "Ok, I will continue from method" first and ensure code blocking is correct.
I think though as long as these remain text generators, we'll be fine. You can't take instructions and build an app alone, you still need to follow them and address any issues, just like how there are articles out there with the same detailed guides. But it'll SERIOUSLY amplify the productivity of anyone who uses this. How this will affect the industry I have no idea, but I hope it'll be for the better and we simply just produce stuff faster, and more reliably.
@@codewithstephen6576 I've unsubbed because of the clickbait titles and thumbnails. It's a shame because the guy appears to be a reasonable programmer in some areas. Hopefully the notifications will stop bringing me back shortly...
how can i adapt to a world where everything, EVERYTHING, i can do a machine can do better and faster?
we're headed towards a dystopia, millions middle class jobs will be cancelled (not just devs), by several of reckless big tech companies with a history of unethical business conduct.
and nobody seems to be concerned
Won’t your career be in danger too? What are the solutions? Cause this just sounds like cheap clickbait.
I love how I'm starting as a jr dev in 10 days and see stuff like this
Welcome to the industry that's been about to make all programmers redundant since the early 80s.
I don't think so, that its gone a happen so quickly, you have still time, but harry up. Finish is about 2-4 years )
Perfectly placed to learn and embrace the new stuff. It is the guys who's experience just got devalued who have to worry.
AI isn't ur enemy here, it's the endless job requirements and needed at least 15+ years experience that will get you :p
TwoMinutesOrLess In 10 years lots of people will loose their job to AI so keep grinding
I have 17 years programming experience. I also pay for the pro version and it's definitely made me a better developer. I'm becoming insanely productive.
Don't lie
20 years here. Delphi, C and Rust developer. CGPT has helped me understand some topics I found complex and has made me a better developer. I just hope it doesn't take my job one day.
I use it mostly for reading library docs and for writing tests, and yeah this thing is scary af
@@thatsalot3577 I also use it for writing unit tests. I usually have to modify them a little bit, but they're basically amazing and save a lot of time.
Same here. But I work too much even with AI generated code
Great presentation Nick.
30 years developer here. The main question for me is how we (humans) will not be eliminated from the creative process, in development, but also generally speaking too.
Obviously ChatGPT 5,6,7,8 ... will solve all our problems. This is not only development "tool", it goes and will go far beyond our comprehension, and that is the scary thing.
We will not be able to understand how this platform "thinks" and generates solution and at some point for the 1st time in human history we will need to watch what this is doing in a similar way that our dog is watching us now....
Such tools should be open sourced and available to human control otherwise 2-3 big tech companies will literally be able to dictate everything on the planet.
prompt engineering is the new job right now. I mean as long as general purpose AI doesn't occur then the big company still need a bunch of Einstein to make it happen, so maybe we still have a chance 🤔
As an AI language model, I can understand your concerns about the increasing role of automation and AI in the creative process. While AI has made significant advances in the fields of image and speech recognition, natural language processing, and decision-making, it is still far from completely replacing human creativity.
In fact, AI and humans can complement each other in many ways. For example, AI can help automate repetitive tasks and provide insights based on large datasets, while humans can bring creativity, intuition, and empathy to the problem-solving process. In other words, AI can be a powerful tool for enhancing human creativity, rather than replacing it.
However, I also understand your concern about the need for transparency and control over AI systems. It is important to ensure that these systems are designed and implemented in an ethical and responsible manner, with clear guidelines and oversight. Open-source platforms can be one way to promote transparency and accountability, as they allow for community collaboration and scrutiny.
Overall, I believe that the future of AI and human creativity will be a collaborative one, where AI serves as a tool for enhancing human creativity rather than replacing it, and where transparency and ethical considerations are paramount.
Yeah these systems will not just replace programmers but they will replace pornstars, onlyfan models, they will create ALL content on the internet at a whim, type what you want and it creates it.
@@user-eo1vk This is a nice and beautiful example of reply from an AI model. But, should we believe what it says? Should we take into account these words? Of course not. At least at this point, my question, should be answered by humans first, because it concerns humans on the first place.
This is EXACTLY what Elon has been warning about for years. If it's privatized, we're done for.
Things I have used it for outside of programming that impressed me. I was rewiring a skidloader and it was able to troubleshoot a 60 year old machine, it was able to tell me what wires went where, resistance values to check if parts were still good, and even tell me factory bolt head sizes so I didn’t have to guess.
I also use it to type out concise instructions for specific models of electronics explained “like I’m 5” for my elderly family members. It was able to walk my grandma through cleaning her print heads and connecting her canon printer to wifi, a lady who has never even used a smartphone.
Being a good developer and learning how to code shouldn't be just about typing a bunch of well structured documents. At it's core, it's problem solving, or should be. I'm a senior developer but I know i don't want to be doing this in my 50s, that's why I've started my own business last year. Think of dev skill as a set of skills to a mean,e.g. running your own business. Put it this way, WHEN these AI tools can do what I do as well I do today, I'll use it to create. It'll be like having a team of 100 developers at my disposal to run my business.
To all junior devs, do not be discouraged by your journey. Coding teaches you how to think, it's a fantastic journey. Use these tools to frog leap your career, not just being a code basher in some office till your 60s.
If coding teaches you how to think you probably should not get into software engineering. Software Engineering requires good thinking and there is not a lot you can do to improve your inteligence.
@@btm1 thats absolutely wrong. Logycal thinking is something that the more you do, the better you get at. Our brain has neuronal plasticity, so it adapts to the environment.
@@btm1 intelligence = knowledge
@@btm1 that's incredibly way off. Here are some facts, not opinions. What we consider intelligence is generally dictated by our neural connectivity. i.e. how well connected your neurons are connected to one another. We can create new neural pathways, that's how you learn stuff. And repeating stuff strengthens the pathways, that's why you improve in something the more you do it.
Yes, some people have higher genetic potential, but you can still massively improve at something mentally and be considered more "intelligent" by simply training. That's why it's highly recommended for old people to keep mentally active.
@@btm1 it's obvious from your statement that you have never truly put years into anything difficult.. And even people with an IQ of 150 can still devote time to do something that they find difficult and grow through the experience...
😀 when you realize that chat GPT can do your work.
😨 when you realize that chat GPT can do your work.
If you use ChatGPT to replace junior developers or any other junior level jobs, who would you have senior level jobs eventually?
good point
ChatGPT, clearly
AI of course, as it gets better it will be able to handle more complex tasks and it doesn't even take a lot of time. ChatGPT was released in Dec 2022 and we have this now. It's only been about 3 months
I think this is the main problem we are facing and we are going to fail to address it hard given human greedy nature. It's not just about developers. It's about every single profession whose entry levels are going to be replaced by AI. Artists, Developers, Copy Writers and the list goes on and on. Those jobs will soon be gone due to cost optimisations and once that happens in x years people to actually handle all those systems will be nowhere to be found.
Looks promising. I will start to worry when he will be able to load at least 1000 lines of legacy code and, based on the description of the task, fix or supplement the functionality.
Give it about a month.
@@stepbackandthink A month is a bit short, no? I would expect premium products to be able to accomplish this next year at the earliest - more likely within the next 3 years.
Edit: Unless you use something like Github Copilot where its only task is to parse the code and provide answers. But something general purpose like how chat gpt is now will take some time before being able to be used like that
It'll be easier to fully replace the legacy code base from scratch using AI.
Would be cool if it could comprehend your whole legacy repository.
The old model actually did that pretty well, this one given that it can take in context documents with 10K+ words in it fine will only be better. Actually considering buying plus, just to see if it can refactor my shitty code from my projects.
Still figuring out how to make the most of it. As an experiment I tried to create moderately complex software dictating the requirements to GPT, but felt it takes more time than if I just coded all by myself. Can still be a useful tool for troubleshooting.
The irony is that it takes an experienced Nick Chapsas to assess whether the output is correct or not. A junior developer would be somewhat blind, and a company executive or manager would be even more blind. You could fire all of your software engineers, but ChatGPT wouldn't help you to build an enterprise software product (vast understatement alert). Writing code is just the unit of work, it's the abstraction layer above that of understanding the business requirements and creating a working, maintainable, performant and scalable solution that is where a software engineer's real value lies. People said that the desk top calculator would make accountants obsolete, but they still exist, because number arithmetic is just the unit of work, it's the abstraction layer above that of financial accounting which is where the real value lies. The calculator just automated the tedious part. If they ever crack general AI that could be another matter.
That accountant and calculator example has no zero relation to AI and its impact. We don't call it "AI" without reason.
By the way, calculator made many companies obsolete. There were many that did manual calc work, accounting firms outsourced much of that work. With invention of calculator all those companies went broke.
You can fire all your software devs and hire one Nick and he builds your product with gpt 4
@@loremipsum3147 lets me honest, you have no idea what software development is
This is awesome, I think I'm going to find ways to utilize this in my everyday job like how you did for example to create test cases for specific things and also create easy api call scripts that could save me 15-30 minutes sifting through docs
Personally I think this is the future, a coding assistant that does provide help.
Not a coding assistant. Just an assistant. In every scope.
I highly suggest you still take the time to understand the code. Writing code is easy. Reading code is a whole other story.
No, your unemployment would be the future.
@@ilovetech8341 that's the point. And the current AIs far far far away from understanding complex things, I mean they don't understand anything to be precise
@@youtubehatesfreespeech2555 Then just create your own company and that should be OK.
Using chatGPT to help me write code for my game and answer questions on generally how to do things. Really improving my mood as now everything seems way more achievable. One questions away and I get at the very least a really good pointer in the right direction. GPT4 is for sure better in programming over 3.5
Do not panic. I've been using ChatGPT and even with version 4 it still needs tons of babysitting and code refactors. You even need to adjust the code as it often does not compile and/or misses bits and bobs. It's a great tool to get some outlines and ideas on how to tackle certain situations, but this is a just TOOL that developers can wield.
Coper. People seem to be willingly ignorant how things are going to progress and how fast.
This would only reassure very senior engineers who are still wilfully ignorant about how fast things are progressing. Juniors are rightly terrified - from a technical coding perspective there's slim chance of them learning faster than AI is progressing in this field. Even trying to do an end-run by focusing on domain knowledge seems doomed.
@@bokkeman123 just learn to work with the tool, that's my point. You'll probably start seeing workplaces offer you chatpgt plus with your resharper license when you join a company.
@@youtubehatesfreespeech2555 Hyper. People seem to be willingly hyper about a progression curve they know 0 about
@@youtubehatesfreespeech2555 Only developers are stupid enough to create somethings that will render them useless in a couple of years. So yes I can fully understand the people that cope.
I’ve been working in c# for a year and I’m happy that I know all of the things you were talking about. I guess I’m considered a senior then? I’ll be using this tool for sure (when being given new projects to work on)
Thank you for the video!
I think it makes sense to also test how good it is now in adding new features/refactoring/fixing bugs in the existing code (since it may save even more time for devs).
Well, all this is quite a straightforward task in my opinion, that indeed the AI should be able to handle easily. And indeed, it might replace junior developers, but the question is - how do we create senior developers if we replace junior developers with AI? And more importantly, it is non-trivial, open-ended tasks, which might even not have an optimal solution, where the human senior developer adds value, not so much with CRUD api programming.
Writing IAC and knowing how to scale an API are not straightfowards tasks at all. Neither are Integration nor performance tests.
It is an interesting question indeed, we already have a huge shortage of Senior Developers!
ChatGPT will just make this shortage even worse in the long run.
that answer to your question is very simple: we don't replace most seniors because the no. of jobs required in the future will continue to drop due to automation of the field.
Slash the # of developers and you slash the training data like stackoverflow used to train these models. How do they train them without massive amounts of training data?
@@nedames3328 that's a problem for the future... that will be pushed out, years away
I am so glad I have retired. When I started (in Australia) there was no internet and I was constantly reinventing the wheel. Work was all about creativity and discovery. Tomorrow it will be all about who can craft the best ChatGPT prompts.
true
yup. just coding equivalent of SEO
You're pretty lucky you didn't have to worry about something replacing you. I'm a CS student and I'm pretty concerned about this
Here is a pro tip if you have chatgpt run out of tokens on script. You copy the last bit of code and write "could you finish writing this code" and it will pick up were it left off. Also this can write G-code for cnc or 3d printers
This is essentially what I originally thought the capabilities of GPT3.5 was because I had seen the capability of what gpt3 had like a year prior with tests people did and when I used it I wasn't that impressed.
GPT4 though...
This seems to have way better context awareness and is now actually worth paying for. If you have used chatGPT in any meaningful capabilities, now is the time to fork over the $20/month.
I asked it to create a non-trivial webpage with api and database connections (you know like any website these days would use) and it did it in under 30 mins with only 2 resets. Copy pasted close to 1200 lines of code and it worked minus 6 easily resolvable bugs that it instantly fixed. The original chatGPT straight up wouldn't have been able to do that unless you spent something close to a dozen hours babying it along.
Well I guess creativity will be the main distinction point when applying for jobs. And knowing design patterns, algorithms and data structures like you were born with them will be a must.
I feel like my job is more figuring out requirements than coding. GPT feels like an updated version of Resharper, not going to change much our job, but accelerate the coding part maybe.
I like your analogy.
You are right that software development is not just about coding, but also about understanding and analyzing requirements, designing solutions, and testing and verifying the software. While AI language models like ChatGPT can help accelerate the coding part of the job, they may not be as useful in other areas such as requirement gathering and analysis.
In fact, AI language models like ChatGPT are not a replacement for human expertise and creativity, but rather a tool that can help augment and accelerate certain aspects of software development. They can help generate code or suggest potential solutions, but they cannot replace the need for human developers to analyze requirements, design solutions, and ensure that the software meets the needs of the users and the business.
Therefore, it's important for developers to develop a range of skills beyond just coding, including the ability to communicate effectively with stakeholders, analyze requirements and design solutions, and ensure that software is tested and validated to meet the needs of the users and the business. Ultimately, it's the combination of human expertise and AI tools that will enable developers to build high-quality software that meets the needs of the users and the business.
And reduce the value also.
Also, Crane on a construction site didn't replace everyone, but now, a single person operating it, is a replacement of hundreds of workers.
Downsizing, more competition, lessened value is what we will face at some point
@@randompoop1565 Did ChatGPT generate that comment?
@@teamdoodz Why don't you ask ChatGPT
As a team lead the main skills I look for in any developer of any level is problem solving skills, clear communication and clean code skills. And more often than not developers really struggle on these (even seniors) and that is what is causing delays and badly designed solutions. So chat gpt is not a threat to those who have or cultivate these skills, but it is a threat for those developers who don't and think writing boilerplate and inefficient/unadapted/unmaintainable lines of code is the only thing they are paid for.
The thing that blew me away was when I asked GPT-3.5 to write an analyzer. It's a pretty niche topic compared to web stuff, but it performs really well. For any kind of complicated stuff it kinda goes off the rails, but the output still saved me a lot of time.
What does it analyze?
I see this as a means to make developers more productive. Granted, it might mean that your boss thinks that because you're getting so much done he might not hire a couple of juniors to help but generally being more productive results in more work eventually funneling your way... Software is never finished.... At least until someone notices it's not being used.
Yes, but all that means is that an organisation needs less programmers overall - so there will be less jobs available.
@@Retrohertz If we wrote software like we did 30 years ago then we would need 100 times more people doing the job today than what they actually do. Work developers have to do have already went down tremendously yet number of developers have only grown. Because ultimately it is the opposite problem. It is not about there being a constant amount of work and x amount of developers. The amount of work changes. We have way more projects than we had in the past and we will have way more projects in the future than we have today. Simply because financial bareer to built new apps will go down significantly so it will make more sense to built more stuff. This means that even if projects will now require 2 people instead of 5, it does not matter if for every such project there are 3 in the future.
Nick, I hope you are going to be OK. Haha. PLEASE keep making videos about Chat GPT. I learn so much each time. :D
Thanks for the "continue" tip. This is amazing. I can do in one day a job of one week before. Since I charge by the hour, I get paid less to produce more. Arg!
My concern isn't losing my current job.
My concern is never being able to get another one.
I heard McDonald's hiring.
Its just a very fast research replacement for Stackoverflow and Google. But what will happen, with Chat GPT if developer blogs and Stackoverflow are dying?
Addons to a Visual Studio or IntelliJ to automatically generate all these code files within a project with just a simple ask what you need. Color code generated code files, so reviewer/developer can go directly to these files to review/tweak the rest of the code. Same kind of addons to ask to generate UnitTests with 100% coverage for the whole project. :)
Only a matter of time.
It's a force multiplier on anyone that knows their task, at least. The extra productivity will ripple through industries.
My problem is that, unlike natural languages, programming languages are highly formalised and not so subject to misinterpretation. It follows that training chatGPT encodes prior art and copyrighted code far more exactly than natural languages. So, what is being generated to a query is much closer to what a paid professional programmer would do...in this case the programmers who supplied their code for training. Now while it is true that most? of the code is open source, I am not sure how I feel about hard won coding techniques used to get an actual paid job are being inferred from public code and used effectively without attribution. Now if the government recognised this and paid me a UBI I would say to chatGPT - go for it - use all my code.
I totally agree with you.
Basically what we are seeing right now: people producing open source code, that actually works, and getting replaced (or at least harmed) by this new tech, The Karma for devs came too fast hahaha lol, the same thing happened to the Artists and most of "we" laughed/ignored that situation, and now, we are here, stuck on the same situation.
I don't give a damn about your closed source 'professional level' code and I hope you're chocking on it. Even if they trained they rearanged and mixed the information from a sea of code. WIthout people sharing you'd prob still be playing with a heat brick that generate pixel art
To my knowledge:
The Legacy model is the original GPT3 which where fine-tuned to more human-like answers, the so-called GPT3.5 (thats why the name).
Then GPT3.5 Turbo (GP3.5 Default), which where released few weeks ago. (1st of march)
my favorite response to GPT is "if i knew everything on the internet i could do this too"
That would not be enough, you would lack the insane ability to reason, calculate, and more than the AI has. It would take you forever to do a little bit of math that the AI does in a millisecond. And your brain wouldn't be able to store all that information either. Let's face it, we're just meat bags becoming irrelevant.
I have decided to directly join as Sr. Developer and am skipping mid/Jr. developer role.
First you get rid of juniors and middles and then suddenly you have no seniors
Or as I call it ... COBOL Syndrome.
Impressive. Although it still needs an senior developer to review the results (as you said) AND foremost it needs a developer that writes the right prompts. :)
Being a software developer will change, without a doubt. A quote from the movie "I Robot" comes to my mind : "That, is the right question!" ... that's what it will be about in the future. You as a developer have an A.I. companion that does the actual work for you, but you need to command it (by asking the right question aka write the right prompts)
Now yes, but if you wait a couple of years what will happen then?
The consequences will be enormous.
I see two scenarios one with mass unemployment i.e everybody go basic income and the wealth is consolidared to a few. But there will be a nisch like the amish that will live in the old economy.
Or second scenario where everybody will be able to create anything. Which will lead to an abundance of goods and services.
Imagine you want to build a car, you can now get the design, drawbacks drawing, the specs and 3d print it.
If you want to make a movie, you can write your own storyline, who should play what role and how they should play it.
I don’t know what the next generation will do.
Ironman's AI assistant.
It depends on the complexity of the job. For most stuff all it requires is a beginner developer, especially if the language is one the developer is familiar with.
@@loot6 I disagree. A beginner developer can't possibly evaluate the result for being actually good or not.
Great point. Just not a new point. If you learned how to search the internet when it came out, if you learned how to ask the right questions, you have been ahead of the curve before this thing.
Well, as an architect, this tool is amazing for rapid prototyping, fast POC generation. If I could somehow convey system design for multiple services and get the scaffolding that would be awesome
I can see few years from now some would try to integrate it to other software.
For me, the feeling is like the world changed yesterday, like the first time I've browsed the internet...
Once that can be integrated into an IDE so that the AI can know the entire context of an existing project, that's where it gets really useful. Then our jobs shift completely towards doing code reviews. And eventually we can skip code completely and our projects will just be sets of use cases that can be used to build a project from scratch every time. Most of the complexity of "software development" will be in maintaining those sets of use cases and making sure no data gets corrupted during updates to a new version of the software.
Github copilot x
Correct
Also replace our projects with my company projects. You'll be working as an AI director in a company with team size of 1. You'll report directly to the CEO and monitor your AI devs each day. Just ensure that your employer does not replace you when chatGPT version 100 is rolled out.
The best part of integrating this into an IDE is that the AI won't HAVE TO know the entire context of an existing project, as long as the programmer is working with enough skill at design patterns to use interfaces and perform things like dependency injection, which will drastically reduce the surface area of the context it will need to program.
TL:DR ChatGPT with Separation of Concerns will reduce the number of credits used to help with programming.
from experience the dev work is so much in the customers understanding and changes in business. which i don't think a bot can do but can only help.
I am right there with you Nick excited, scared, terrified, super optimistic.
So many ramifications..
I guess more developers will move from being code creators to code reviewers and "fine-tuners", which is fine, you will still need the knowledge and experience, there is no skipping or losing anything
It's not the 100% elimination of the need for humans. It's the 80% elimination of humans not needed because you need 1 or 2 people for the 10 you need now and those 1 or 2 people are the elite of the pack. The other 8 or 9 out of 10 people then have to become plumbers or working retail at Sacks 5th Avenue.
@@dwcola yea thats what most people didnt understand. AI not gonna eliminate jobs. It ll reduce it. And AI is not just another tool . Its an advanced intelligence that have potential to replace humans as the most intelligent species on the planet
@@dwcola this is exactly why it is so scary and its not just coding it is the same in every knowledge industry, moreover its still time to reach end of 2023 may be next version will cross all the barriers
How will new developers get skills to do code reviews and fine tuning if they don't get experience?
@@RonaldDraxer-rb4qm The LLM version of AI isn't the advanced intelligence kind. It is an industrial revolution type of efficiency gain in programming, but it can't advance the art. I've been playing with it and it can't even really understand working code when it's complex enough. My test was trying to get it to generate a working unit test. It could not, even as I deconstructed the task. You still need a person to understand what it is doing. Spitting out the unit test framework and saving a ton of busy work was it's forte.
I've been programming for 45+ years. This is scary and exciting. It's similar but so much more than when SQL fist appeared and even more than when desktops first hit our offices! Embrace it.
Cool stuff, however, I would like to see how it fairs with less "book like" questions and more real world weird requirements and constraints people have. Especially with legacy code. If it can handle that, then we are talking, if no, this is nice co-pilot to help me skip the boilerplate in some cases.
Nice video Nick, I hope to meet you at NDC Oslo and discuss applications of GPT to do coding tasks. Some of the code in my talk is generated by GPT, I think all developers should embrace this.
I think that gpt4 is pretty amazing, that being said it makes some pretty crazy errors (although it is better than gpt3). I was using it here a couple of days ago for something, it produced some code and it messed it up and I corrected it and moved on. The problem with where it messed up is it would have been nearly "invisible" to someone who did not know better. This tool is going to end up making a lot of subtle mistakes. What I think will happen is many companies will try to get by "on the cheap" with junior developers rather than senior developers and those junior developers will accept A LOT of mistakes into the code base. Also as more and more code ends up being the "training set" for these tools that was generated itself we will start seeing some really amazing bad problems. A few (very stupid) companies may even try to let project managers use this to produce code. That is where it will get incredibly bad. I think that we will have to use these tools in order to keep up but it really is going to be like stack overflow on steroids. Errors will still show up and our job will end up more to catch super bad errors before they get integrated into large code bases. What I am actually afraid of more than anything is having to clean up the mess this is going to cause.
I think companies would rather try and make do with a few SENIOR devs and rather replace the juniors with AI tools.
"stack overflow on steroids" is my current view as well. By looking at my daily work, the total number of line code I produced is not a big number. Most of time is to understand existing code base in different components/subsystems in a big distributed system, and identify what is wrong or missing. I don't see how ChatGPT can help me for that because even if I try to explain it my co-worker, it might require white board 😪
i think you speak too soon...look we r just a week into this...comment from september 2024...about LLm's....they will be kinda better then
Reasonable take which I agree with.
Reasonable take which I agree with.
One thing to keep in mind - I only ever see ChatGPT tested at creating new projects or writing new code. But most software development work is changing existing codebase. It would be very hard end time consuming to explain to ChatGPT how your existing codebase looks like, what are the requirements and limitations etc. So it looks impresive but it's mostly usefull for boilerplate setup. Be honest with yourself now - how often in your real job do you actually set up new things?
It was also trained using uploaded code, so because commercial code was not used as training model it will do really bad in real-life commercial stuff.
It's true, but imho it's only a matter of time that it can connect to a codebase and understand it instantly, then a refactor that would cost 1 year of dev time could be done in 10mins, and adding features would be easy as connecting it to the Ticket manager so it can read and execute the tickets.
And how often do we write overly-simplified Movie databases? :)
Now imagine an IDE plugin that's a hybrid of GitHub Copilot and ChatGPT
I’ve actually used it to provide legacy code and have it review it and refactor it. It is really good at it
What I see is it is a great tool for learning. It has the knowledge of an expert about a lot of disciplines. And it can automate generic and trivial works, but at the moment, cannot analyze your project or a real scenario. It's a cool, powerful and helpfull tool for taking decissions, but the humans are who finally take decissions, and modify or customize the recomendations. I see the tool as a Stackoverflow with steroids.
Junior developer today, senior developer tomorrow. I've been saying for years that coding of programming if you prefer is a mechanical task well suited to automation. What is not suitable for automation is requirements, that is specifying in detail the software functionality - what my grandmother used to call the gizinta and gizouta.
I agree that the role of a developer can evolve over time, from a junior developer to a senior developer, as they gain more experience and expertise in the field. As for the question of automation in software development, it's true that coding can be seen as a mechanical task that is well-suited for automation.
However, as you pointed out, requirements gathering and analysis is not as easily automated. This is because requirements gathering involves understanding the needs of the users and the business, and translating those needs into specific software functionality. This requires a deep understanding of the problem domain and the ability to communicate effectively with stakeholders.
While AI language models like ChatGPT can assist with generating code, they are not as well-suited for requirements gathering and analysis. This task requires human expertise and creativity, as well as the ability to communicate effectively with others.
Therefore, while automation can certainly help accelerate certain aspects of software development, it is unlikely to completely replace the need for human developers. Instead, it will likely augment and enhance their abilities, enabling them to work more efficiently and effectively in collaboration with AI tools.
@@randompoop1565 thanks for the response chatgpt
> What is not suitable for automation is requirements, that is specifying in detail the software functionality
This is just an "idea guy" thing that can be done by the client or business people and not require a developer. All the technical work is going to go away.
2023/2024 we automate the junior developers, 2025 the senior, by 2028 even the idea guy is gone.
@@MiraPloy I still think we need technical people. Keep in mind it doesn't actually do those things, it's just generating text, and while detailed and comprehensive, I doubt non-techincal people would be able to follow. We have wiki's, guides, and articles that we use already which are the same, but with a much narrower scope. I think in order for this to be properly effective you need both the model and the person who is also able to learn from it. The way I think about it, is this is a tool that turns juniors into seniors.
@@drmonkeys852 I think you can trivially imagine an app that takes the AI output, have AI create tests, and then automatically test and compile it.
Sorry mate there is no saving humanity, it will all get automated.
I wonder if it'll ever get updated past 2021, the coding world is changing fast and like he said in the video "We don't do that in c# anymore". It would be great if it can provide more up-to-date techniques.
Scary/funny that you asked this 4 days ago and already the new Plugins are in play to fetch live data from the internet. Any online doc for a language will be found and parsed...
“This is gonna save me so much time!” Yeah, I think you’ll have a lot of free time pretty soon 😂
Already waiting for a basic income salary from gov :D
@@scoff7032 ‘Fraid so, bud.
I had an interesting experience with ChatGPT 3.5 yesterday. I was trying to get it to write a very simple shader for Unity. It seemed to know what it was doing at first, but kept writing the code incorrectly. Every time I sent it the error message, it correctly understood what error it made, but it then wrongly applied a "fix" to the shader. I basically spent half an hour repeatedly pasting its shader script, reporting the error, pasting its fixed script, reporting the error, etc. and we just went in an endless loop where it could not give me a shader that didn't have at least 1 erroneous line.
I wonder if ChatGPT 4.0 would be less... I don't know, incompetently looping if I asked it to solve the same simple problem.
For anyone who has Plus and wants to try it, all I asked it to do was to give me a simple shader based on HDRP Lit that has a color texture input and a noise texture input, and to make the noise texture displace the vertices. After over an hour leading it around, it was never capable of completing the task. I'd be interested to see what 4.0 can do better.
True, it may help with productivity, but I still think is important to have this juniors and mids getting familiar with the stuff even if it takes them longer. Learning and getting pro takes time. Its the only way to make sure we keep having great seniors in the future.
When my father programmed in the 70s, he had to use punch cards. Then in the 80s, it was C, Turbo Pascal, and Borland C++; in the 90s we slowly got into GUI design with Visual C++ and Visual Basic; then, in the 2000s, modern IDE came, and later we had Stackoverflow, cloud-first development,... Software development has changed a lot over the past decades, and it will change rapidly in the next few years. Development of a complex system will be more like drag and drop boxes of logic, where the IDE will help you to design the details of the integration and adjust little logic. Software development will move more and more toward system development, and details will be more abstracted away. So, software development will move from coding to more and more of a solution and system design.
I agree that software development has undergone a lot of changes over the past few decades, and it is likely to continue evolving rapidly in the years to come. As you mentioned, the development of complex systems is likely to become more focused on solution and system design, with coding becoming more abstracted away.
This shift is already happening in some areas of software development, with the rise of low-code and no-code platforms that allow non-technical users to build software solutions using drag-and-drop interfaces and pre-built components. These platforms abstract away much of the coding details, allowing users to focus more on the solution and system design.
However, it's worth noting that even as software development becomes more focused on solution and system design, there will still be a need for developers with deep technical expertise to build and maintain the underlying infrastructure and components. As with any technological shift, there will be new opportunities for developers to learn and adapt their skills to meet the changing needs of the industry.
Ultimately, the future of software development is likely to be shaped by a range of factors, including technological advancements, changing business needs, and evolving user expectations. Developers will need to stay current with the latest trends and technologies, and continue developing their skills and expertise in order to remain competitive in the job market.
@@randompoop1565 Stop sharing that ChatGPT generic bullshit.
@@jendabekCZ Why not? Sooner or later, all your responses will be from chatgpt…
@@randompoop1565 Haha, why do you think so?
Floor sort!
I don’t think chatgpt can replace a dev. I literally use it to build a big side project I build, sometimes chat gpt spits out stuff which doesn’t make sense, sometimes it suggests functions which doesn’t exist in some libs and stuff
Dude, it's been 2 days 😂. What are you talking about??
that is because this tech is in its infancy. Tech evolves pretty fast, I'm a student and it already knows more and does more than most students I know. 5 years from now it will be drastically improved.
@@d_pratik1 this
Looks like Within few minutes we convert an idea to real application. I am really blown away 🤯. You are amazing Nick
Maybe the first time I watch one of Nick's vids in 1x instead of 2x cause of the sheer fascination and shock 😅
A new wave will come: Instructional programming.
I already use ChatGPT regulary to help me in my daily work and I agree that the new model is very, very impressive.
But the real task of developers is to work with the requirements we get from the stakeholders.
What I mean by that .... requirements are never a complete definition of the result. Think of it - if a requirement would be a complete definition, it would already be the actual program!
So maybe in the future the job of us would shift to write the "programs" in a more natural language, but it is still translating all the babbling from the stakeholders to some unambigious form ...
and we will be much faster in putting it in an actual running form or refine it in an iterative way.But it will not replace us. At least not all of us ;-) ;-)
Requirement defining is an art.
There are a variety of requirements that range in terms of complexity.
Hence a product manager or BA is much more relevant than the coder (am aware am naming it coder, because in most of the backend jobs in India it's just coding that they do, no thinking), this can replace crazy bunch
That is what the product manager does.
Well said! It will not replace us, but slowly as our jobs get easier, the pay might get lower. The thing we will be paid for is for clients to have a human to interact with. It will take decades for AI to be considered trustworthy.
That is true. Context lenght is really huge compared to 3.5, after an hour of chatting, it can still recall the initial prompt amd make a summary of everything. At the end of the session, you can ask it to generate a prompt that collects all reguirements from numerous separate prompts and answers. I'm blown away too. I wish they revealed how to keep context alive so well in API version too. Im pretty sure they use some sophistocated way of "compressing" chats...
A chat bot demo app that leveraged the API took the approach of just posting the ever-expanding complete chat with every request while asking for the next response. Not only that, but they also included the domain-specific background info to inform the bot's response.
@@bokkeman123 In Open AI GPT chat models context cant be ever expanding. Messages must be trunkated to fit maximum token size, 4k, 8k or 32k. I havent seen any doc which explains how this limitation is handled in ChatGPT. Logical explanation would be to summarize text which condenses it to a smaller space. And keep summarizing it when new messages come. But to implement this would require still more API calls and increase of costs.
@@MarkoTManninen You might find this interesting - rather build your own model trained on your domain data, or use "embeddings" to enhance an OpenAI model th-cam.com/video/veV2I-NEjaM/w-d-xo.html
AI is replacing everything, honestly, in a few years time you're gonna be able to put a few statements in and get a whole movie out of it - this is the ultimate pandora's box. With regards to dev, in future you might not have several junior devs, you'll have AI administrators. The day you can upload a legacy code repository to an AI and ask it to re-write it all, that's the day it is over for most developers
While it's true that AI is rapidly advancing and has the potential to automate many tasks, including software development, it's important to keep in mind that AI is not a panacea and there are still many tasks that require human expertise and creativity.
For example, while AI language models like ChatGPT can generate code, they may not be able to understand the full context of a project, including the business requirements, the technical constraints, and the user needs. They may also struggle with tasks that require creativity, such as designing a user interface or developing a novel algorithm.
Furthermore, it's worth noting that the development process is not just about writing code, but also about analyzing requirements, designing solutions, testing and verifying the software, and collaborating with other team members. These tasks require a range of skills that are not easily replicated by AI.
That being said, it's certainly possible that AI will play an increasingly important role in software development in the future, and developers may need to adapt and learn new skills in order to remain competitive. However, it's unlikely that AI will completely replace human developers, and there will always be a need for human expertise and creativity in the software development process.
@@randompoop1565 Did you really have to use chatgpt to write that text lol
AI admin, that's a good idea
@@tvojtatko123 he really thought that removing the “in conclusion” paragraph would make it more low key 😂
Well, if AI is indeed replacing everything, I guess we're all just a few lines of code away from getting six-pack abs and a mansion on Mars, am I right? 😄 But seriously, I can already see the movie titles now: "Artificial Intelligence Saves Christmas" or "The AI Who Loved Me." Move over, Hollywood! We've got algorithms in the director's chair!
As for the devs, who needs human developers when you can have a legion of AI administrators? I mean, imagine the water cooler conversations at work: "Hey Siri, did you catch the latest episode of 'Westworld'? Oh, wait, you are the Westworld!" 🤖
Oh and by the way, it's a gpt-4 answer
It would be interesting to understand how the changes and improvements of GPT-4 actually work under the hood. However, the 80+ page apparent "research" paper that was released only states that it "works a lot better," but has absolutely no information on how the new features actually function. It feels more like a sales and marketing material than an actual research paper. OpenAI is being somewhat disingenuous, which is similar to ChatGPT's tendency to make believe things that are untrue at times. There are also questionable safety concerns that are restrictive to others who are less privileged. While it is understandable that a company wants to make money, they should not claim to be "open" or open-source when they are not. OpenAI should be more transparent or be called out for this deceptive tactic, as demonstrated in the little poem it ironically made with its release.
I think that this may ultimately change development about as much as the internet (being able to Google stuff) changed development.
So, yes, a lot. But there will still be professionals with coding knowledge involved in the creation of software.
*You* identified the limitations and issues of the answers.
I would be interested to see if the AI can be self-aware enough to list and evaluate the respective limitations in context.
yeah you have to remember what ia gives us is what we humans gave him. If in the next years, nobody shares any code or help each other because we can all rely on IA, soon enough IA will have outdated data and no new data to consume. Using IA is killing IA
Interesting, I have tested it today, and had so many problem, it messed up GCP API with AWS API, so basically it created some unusable code. And my questions were pretty simple. I have spent more time on fixing the API usage and finding the correct one, that I would probably spent on reading the documentation. Maybe I did try it on GPT-3 ;) I just tried some generic questions on trial free version.
I suppose I still need to learn how to work with it.
Nice knowing y'all. What do you guys plan to do now that your job has been rendered obsolete?
That example you gave about GitHub raising issues and having GPT-4 fix them: you can kind of see the progression of this. A self-repairing, self-improving system. Writing enterprise-level applications through prompt engineering doesn't seem far away.
A fun experiment would be to write a blank-slate app, and have users write prompts to add functionality to it. Code which the app itself would implement into itself. Do you see where I'm going with this?
Then the CEO can just fire everyone and run the entire company himself. Heck... maybe the AI can run its own business.
@@ilovetech8341 yeah the idea would be that using prompts you could make an application write itself...and improve itself over time. Maybe use user feedback to implement its own features, feed any errors back into itself to fix, etc.
The same promise was made when Cobol came out in the 60s.
"Everyone can write computer programs now because it's just simple english"
Turns out they couldn't because you still need to know how to properly define the requirements and algorithms for the software.
Unless clients suddenly learn how to absolutely exactly define their needs in software, AI isn't going to take over software development anytime soon.
It's just another very helpful tool for the software engineers.
But it doesn't replace them in any way.
I did this with chatgpt3 to build a tic-tac-toe game. It built something functional with a gui and mouse support, tests, and documentation in under an hour.
@@Youcef8830 yea... until you have to make one little change and it takes days because you never understood the code to begin with.
I literally almost cried. 4 years wasted! Why even keep studying? 😭🤬. I didn't even know what the heck you were talking about IRC LI... whatever. But it did...just wow. I just don't know anymore.
Your time is definitely not wasted. The world needs devs to go through the junior -> intermediate -> senior steps still because it's your creativity that cannot be replaced. You're not a robot.
@@MaxQuagliotto creativity to write shitty workarounds lol
It's definitely not time wasted my friend. Keep your spirits up, this is just a tool, like Google is. The tool needs to be used by programmers, you're not going to see some business man code up an application using this tool, you'll see developers doing that.
I tried using it today to help me write a console application to reproduce a bug I had with an audio API. And it was still like pulling teeth to get it to do what I needed. And in the end i still ended up rewriting the code and refactoring it into a more sane project structure.
Was that version 4?
ChatGPT will now require a different skill set. The ability to use language to specify requirements very precisely is of utmost importance. You'd be surprised at the sophisticated things ChatGPT can code, once you get the knack of communicating with it properly.
@@Steve-xh3by - "pretty please with sugar on top"
I don't know guys, tried to use it at my job and in two weeks it helped me only once in simple task, chat don't know the context, legacy app limitations business rules etc. However sometimes it save time and some responses are really impressive
It's amazing how these technologies are improving day by day. And terrify too
Yet stopping progress is equally terrifying. We can only move forwards and hope for the best and deal with problems as they come.
I think you need to get started on a course to teach us how to manage our new ChatGPT4 employees. I don't see this a job ending moment. This is the moment when the scope of the software which small teams can write, explodes out to a previously unreachable scale.
Completely agree with this. Using Chat-GPT is like giving developers access to a more senior developer that can assist them all day, every day at the click of a button. That more senior developer isn't always right, but they can definitely help you solve problems faster than doing it alone.
As for the job industry. I feel it’s a matter of time before cloud computing AI is being fed tasks through individual contracts/companies instead of physical developers. At this point the training data will be isolated and privately trained for that contract/company. Any hint of that is the beginning of a interesting period.
What I am interested in is how big of a code base GPT-4 can handle. Because the examples we provide it with are small maybe a few 100 lines of code.
Up to 25k words worth of context atm
As time passes and as more of us adopt ChatGPT for developing, do you think the Github repositories feeding data to future versions of ChatGPT will become more and more alike?
I'm wondering if ChatGPT could become "ChatGPT-biased" by influencing future code.
Good point - coding will become less and less 'opinionated' in this way, as more GPT code is fed back into the knowledgebase. Developers will be left with coding completely novel solutions, which will in turn be added to the GPT knowledgebase and homogenized
Thats a good point
This is so scary xD As someone still in school deciding what I want to do as a profession, I'm worried anything I decide upon will be replaced by such AI, in a few years, I see it as a totally real scenario.
Yes, prostitution might still be a thing. And plumber.
Dude, use this to build projects that matter. You have knowledge of how things work, now use gpt to expedite those projects and build things that actually matter. This is far better than being a grunt worker
@@Mosern1977 ever seen ex machina ? Prostitution will be obsolete too one day
@@Mosern1977 Not with sex robots, and labor robots that will soon be in the rise.
@@wantedsavage7776 - I for one welcome our AI overlords.
in ChatGbBT conversation
question : ai is more likely to replace front end or backend development
AI is unlikely to replace either front-end or back-end development completely, but it may have an impact on some aspects of these roles.
In terms of front-end development, AI can assist with tasks such as automating layout design, generating code, and optimizing website performance. For example, AI can be used to analyze user behavior and generate personalized layouts that improve user engagement.
As for back-end development, AI can be used to improve the efficiency of processes such as data analysis and machine learning. For example, AI algorithms can be used to process large amounts of data and make predictions based on that data.
However, AI is unlikely to completely replace human front-end or back-end developers, as these roles require creativity, problem-solving skills, and a deep understanding of user needs and business goals. While AI can assist with certain aspects of development, it cannot fully replicate the human expertise and judgement required to create effective, user-centered applications.
Also : so basically it will reduce junior developer roles
It's possible that the integration of AI into development processes could lead to a reduction in some junior developer roles that involve more repetitive or lower-level tasks. However, AI is not a substitute for the skills and experience of human developers, and it is likely that new roles will emerge that require a combination of both AI expertise and human creativity.
In addition, as AI becomes more prevalent in development, it may actually increase the demand for skilled developers who can effectively integrate and manage AI systems within their applications. AI can bring significant benefits to development, but it still requires human oversight and management to ensure it is used effectively and ethically. So, while it's possible that some junior developer roles may be affected by the integration of AI, there are also opportunities for new roles and growth in the industry as a whole.
--------------------------
We don't have to panic as developers. we just need to understand how we can benefit for this amazing technology and use it to help us. Sure many things will change and those who are not willing to adapt will find no place in the new world of AI.
You mentioned about this not making our jobs redundant, I agree, it had to be trained on something, if we stop coding and making advances the model will stop too, so we would hit a bottleneck.
IMO realizing this will focus everyone on true innovation: phantasizing of things all of humanity hasn’t yet thought about.
The takeaway: it takes a skilled and experienced programmer to ask the right questions and write the current prompts….
I’ve been a developer for more than 45 years and am soon retiring. I think my exit is timed just right. These tools will be used by very experienced developers. There will be no more need for average developers in te near future. It will enhance quality and productivity but make a lot of developers obsolete. We feared this for many years and until now this was never realized but this is a whole different ballgame. Does the future look bright? For employers I would say, yes. For developers I would say, maybe. When these kind of AI’s start building new AI’s themselves the outcome will become exponentially better and outperform all of us. But remember, these models are not smarter than what they can learn and they still learn from us.
Capitalists are getting the upper hand, why share profits if they can have it all especially when society is there to cover their losses. I bet, the latest layoff frenzy in SV is definitely no coincidences and it won't be the last.
It sucks for me since I was just thinking about becoming a web developer.. I was so inspired like a 4 year old waiting in the line to experience it all. So perhaps I would make a pass and do something else? :(
It will learn from us until it does not need to. That’s when Houston has a problem.
@@TheMugenVideos Don't give up so easily. Jump on the train and see where it brings you.
@@TheMugenVideos There is nowhere else to go but whatever manual labour has yet to be automated.
Nick, this is just the months after the previous model. How does your velocity improvement compare?
Finally! A better source for copy&paste code that stackoverflow
Generating a basic crud app is not impressive, you can google the same stuff.
It would be impressive if:
You take a big project with messy code, analyze it and implement a new business feature without damaging your employer's business. And also make it testable and future-proof. That would be really magnificent.
It would be even better if you could remove the middle man (the companies) and have the AI generate to the client directly in real-time whatever they need.
I don't think this will "replace" anyone, it will become a resource for developers to use, just like stack overflow or any other repository of information. It will help to make us more efficient and allow us to focus on larger problems dealing with the specific domain of our application.
Agreed - Whilst you *could* replace juniors in the short-term, today's juniors are tomorrow's mid-level and seniors. Experienced developers have to cut their teeth somewhere. I can see a time where this becomes the go-to place for information, surplanting Stack overflow. It would also be good for reducing the load of senior devleopers when mentoring juniors.
Agree 100%
Couldn’t agree more
It won't replace (for now) but over time it will significantly reduce the need for many junior positions. Considering how fast AI tech is growing, next 10 years we will witness complete shift in this field.
I see programming field suffering from "COBOL syndrome". It's still there but only few know how to work with it (only the most skilled ones).
When you have such powerful AI "tools" many young people will be demotivated to pursue Computer Science degrees. Think how much you have to learn and all the free time you have to sacrifice. And for what ? Just so one day you can become obsolete ?
Honestly, if I had to choose today I think I would rather go into medicine or something where automation has limited impact. I am (fortunately) retiring after 2 decades in telecom engineering and SW developing. My advice to all CS juniors; be smart and avoid CS like a plague.
When a business gets to the point where they would normally consider hiring a new junior developer, but now they can put that off because their current developer can use AI to increase their productivity, that junior developer has effectively been replaced.
Thanks for the video, Nick. Did you try the new Bing Chat and compare it with GPT-4? Just curious about the differences.
I asked Bing Chat the same question NIck started with at the beginning of this ChatGPT session. And it didn't produce any code, it just gave me some links to where I can find more info. Supposedly, Bing Chat is built on GPT-4 but the responses were *not* the same at all.
@@ChristofJans yes, Bing Chat looks weaker
Depends.....if you try to use some update Python modules GPT fails. You have to do a lot of fixes to have the code working in the end.
Even if AI were to replace developers there will still be work for us to do. As Nick said, it will take someone behind the wheel to guide it. Probably our jobs will just evolve.
but they wont need all of you to do the work. It will be just a few senior programmer who will replace the rest
Sorry to brake it to you but that's wishful thinking. Business will remove a lot of people. Senior developers will be the ones taking the wheel. People that are currently doing Uni are going to be outdated if it keeps evolving this fast.
Sure it's a great tool that can help us a lot, but it's also our enemy.
Copium
Lol sure
You are right, but companies will need far fewer developers. It's like farming - what used to take 100 men all day to harvest now takes 1 dude with a tractor a couple hours.
I think you nailed the analysis - amazing responses BUT yes it needs to be reviewed and tweaked. an incredible bootstrap!
additionally would like to add though : remember it is delayed learning and relies on our own industry of developers and testers to generate new content posted to public Internet otherwise it won’t learn new technologies and best practices
Machines can learn how to code direct to the compiler, ditching any programming languages so... I don't think they care about "best practices" if there isn't any practice at all.
But No Code/Low Code already do this and here we are.
It's awesome productivity tool, but to say it will replace us is really funny. I run a company having over 50+ developers and myself being one of them, I have been with .NET since 2002. Majority work that comes is not writing new code but changing it and there are multiple factors at play. Single wrong update can screw business completely. Sure AI will be great productivity boost and may reduce the headcount required, but it will never be a replacement. I think, with AI helping we can create even better things with faster turnaround, and will create more employment.
Congratulations 🎉🎉🎉 I am now become a Developer. 😊
Another good day to start learning farming
No matter how funny it sounds, it's not.
I am retiring from CS after 20 years in the field and last year I bought nice piece of land to grow some fruit and vegetables.
@@SimpMcSimpy There is not a single chance I retiring as well, but I ain't fighting it either. Am adding it to my workflow
Can't wait till this replaces me so I can focus on my hazelnut farm
One day I spent 2 hours and prompted for a TLS file sharing server & client in C#, tcplistener, with authentication, file listing, downloading / uploading and all stats on console. Works perfectly. I dont know what to think about that, other than replacing the terrible FTP server class from 2004. I learned C# over a year from scratch, so this is revealing stream and sslstream methods I have never seen before, let alone in the namespace decompilation view. Not bad! Threw that gem into a secure ZIP file and locked and loaded. Handles 500GB files no problem, buffering and many optimizations. Thanks GPT lol
The trick is to review all of the generated code and learn each method. I was up to speed within the hour! I love it.
Accurate and concise requests and follow ups will get you there quicker.
Also the lack or outdated knowledge is a bummer, and you have to do refactoring but getting proof of concept and research and dev is much faster. But concise and accurate requests .. gets you there very quick
chatGPT made me thinking that my future job is working at McDonalds or some sh*t
nah, that will be robots too
working at McDonalds is already unncessary, but we haven't adjusted yet. Delivering food on a assembly line -> Easy. Ordering and payment -> Already done. All that's left is security and maintance.
Maybe instead of making videos on how “developers will list their jobs” , you should make more videos on about how developers to get a job?…
This is the end of the McJobs and maybe finally the golden age of the Seniors who can interrogate in English GPT-4 and the following ones. There will always be a need for seniors capable of interrogating AIs and supervising the answers, or even modifying them. But for the juniors it is done, we can do without them... And in this case, how will the juniors become the next seniors if they are no longer bankable on the job market? It raises a lot of questions and it's very exciting (even if it's also a bit worrying...)!
I started using it for PowerShell, something I'm already very good with. GPT 3 and 4 instantly made me more productive as an IT engineer. You need to know what you're doing to an extent because it does make mistakes which can potentialy cause it not to run or even take down your domain. but with the right wording it's insanely powerful. I started using it for python and c# as well because why not?
The scary thing as you pointed out is, that companies are indeed opportunistic. It might not be impossible, but all of a sudden become much harder for junior-devs to dive into a job or to receive a good payment, since they need to compete against an AI, that does the same job (and more) for just a few bucks per month. So even if it is right, that you'll still need some seniors to review the code I see kind of a "gap" that needs to be filled to get there. I mean yeah, you might still need some junios, but much less since a fairly good amount of their work can be done bei ChatGPT. I'm not only scared but panicking a bit about this.
I think the way we humans address information and knowlegde is about to change substantially - I see ChatGPT as a calculator - its a tool that you need to learn how to use, if you are a junior, you should be able to hit CHATGPT, get a general idea about what you are building, and tailor your questions. Yes, you will be fighting against AI, but you also have AI on your side to speed up and guide your growth. I wish 15 years ago, when I was looking online for how to do some tasks or a specific explanation for a API was badly written, and lacked in examples. WIth ChatGPT, you can just ask it, and it cures the resources for you.
I understand your concerns about the potential impact of AI language models like ChatGPT on the job market for junior developers. While it's true that AI language models can generate code and automate certain tasks, it's important to remember that software development is a complex and multifaceted process that involves more than just writing code. developers bring a range of skills and abilities to the table, including problem-solving, creativity, domain knowledge, and the ability to work collaboratively with others. These skills are not easily replicated by AI, and they will continue to be in demand in the future.
Furthermore, AI language models like ChatGPT are not a panacea for all software development tasks. They may be helpful in generating boilerplate code or automating repetitive tasks, but they are unlikely to be able to handle more complex development tasks that require a deep understanding of the business domain and the problem being solved.
Ultimately, the impact of AI on the job market for junior developers will depend on a range of factors, including the pace of technological change, the availability of training and education programs, and the evolving needs of businesses and industries. It's important for aspiring developers to stay current with the latest technologies and to continue developing their skills and abilities in order to remain competitive in the job market.
Payment should not be GOOD or bad..payment have to be adequate for the value worker ceates. If your code made 10 000 euros in value - it's not logical to pay you 20 000. More likely it's normal to get 3000.
The fact that you will need less juniors is perfectly logical. Automation has always lead to more jobs lost than created so it makes sense that in the long run you will need less and less programmers.
@@btm1 It'd be more accurate to say you'll need fewer programmers per project, or per line of code produced. Often times automation massively increases productivity, and you just have way more *stuff* being produced, rarely the same amount of stuff but with fewer workers.