SAY GOODBYE to Coding Jobs! AGI is Taking Over! (Should You Learn To Code 2024)

So there was a recent clip that I probably should have covered already, but I just didn't because I was busy. But the point is, is that there was a discussion about whether or not you should learn to code. Jenseng Huang, the Nvidia CEO, basically said in a clip that learning to code is pretty much pointless because we're going to be coding in a natural language. I'm going to play the clip and I'm going to bring up some interesting points that I don't think have been discussed because my perspective, I believe, is a little bit different. I know that a lot of people who are software developers have been discussing this, but aside someone that's been paying attention to the AI space, I think I can introduce a fresh narrative. So take a listen to this and then I'm going to give you all the points on why I think there's a lot more that people aren't thinking about.

I want to say something, and it's, it's going to sound completely opposite of what people feel. Over the course of the last 10 years, 15 years, almost everybody who sits on the stage like this would tell you it is vital that your children learn computer science. Everybody should learn how to program. And in fact, it's almost exactly the opposite.

It's.

It is our job to create computing technology such that nobody has to program and that the programming language is human. Everybody in the world is now a programmer. This is the miracle of artificial intelligence. The countries, the people that understand how to solve a domain problem in digital biology, or in education of young people, or in manufacturing or in farming. Those people who understand domain expertise now can utilize technology that is readily available to you. You now have a computer that will do what you tell it to do. It is vital that we upskill everyone. And the upskilling process, I believe, will be delightful, surprising.

So, yeah, this was quite a controversial statement, and I'm pretty sure he knows that it was a controversial statement, as he said, you know, everyone on stage would tell you something else, but I think he has a point. Okay. And I would argue that there are points on both sides of the argument, and I'm going to present both and then I'm going to give my opinion, because I would argue that for both sides, it's so convincing, but there is one that does take it slightly in the edge. So at first I'm going to give you guys reasons as to why you shouldn't be coding and why you shouldn't learn to code. And this is just if you're starting out as your career or if you're someone that's like, you know, maybe you're 18 and this is why you shouldn't learn to code. You these are just the arguments I'm going to make arguments against these. So just listen to these first. Now, essentially, one of the things that I guess you may have missed was AlphaCode 3. So with the Gemini 1.5 Pro release, Gemini Ultra, whatever you want to call it, Gemini released a series of models. Now one of the models that was kind of hidden. And I know this because I made a video on AlphaCode 2 and the problem was is that it didn't get as much coverage, the video didn't get as much views, but but it did have a significant impact for the future of coding that many people may have missed. So AlphaCode 2 was something that could code better than 87% of people. And they are talking about this as a future system, which is why I'm wondering what about Alpha Code 3, which could potentially revolutionize the industry. So take a look at this short clip from Gemini 1.5 where they talk about the innovation in coding expertise and how they expect the future to change. And this is something I really haven't seen enough people talking about.

We built Gemini from the ground up to be natively multimodal, including something quite important for both of programming code. Gemini is able to consistently understand, explain and generate code that is correct and well written in most programming languages. That includes Python, Java, C and Go. It substantially improves coding abilities over previous Palm 2 models. From a benchmark around 200 programming functions in Python, it consistently solves about 75% of them in the first try, versus around 45% on Palm 2. If you allow Gemini to check and repair its own answers, this number jumps to over 90%, which is a huge step forward. It can help you create and prototype new ideas in seconds. Let's give it a try. I really like trains, and if I wanted to create a trains spotting location web app, I can simply ask and get a working prototype in less than a minute. While the code isn't perfect, it's really helpful to have a first draft. Gemini on its own has the ability to transform software development as we understand it, but it can also be deployed as a key component of more sophisticated systems.

Gemini is great at coding, but we've been able to take it even further, creating a specialized version that performs remarkably well at competitive programming. Now why do we care about competitive programming? Well, it is one of the ultimate litmus tests of algorithmic coding abilities. So you have thousands of Talented programmers from all over the world that come together to compete and try to solve incredibly complex problems that require not only coding, but also math and reasoning. Two years ago, we presented AlphaCode and it was the first AI system that could compete with roughly at the level of the average human competitor. Today, I'm delighted to introduce AlphaCude 2, a new and enhanced system with massively improved performance, powered by Gemini. When we evaluate AlphaCode 2 on the same platform as the original AlphaCode, we solve almost twice as many problems. While AlphaCode broke through the top half of human competitors, on average, we estimate that AlphaCode 2 performs better than 85% of competition participants. Let's have a look at our model in action on one of the hardest problems that we faced. And I say hard because in the original contest in which the problem appeared, less than 0.2% of participants actually solved it.

So I won't bore you guys with the entire clip, but like I said before, I know that a lot of people did miss this. And the point here is that if AlphaCode 2 was this good, I mean, how is AlphaCode 3 going to be? Just try and think about that for the future, because as you all know, these systems get better. I wouldn't say maybe every year, but every, you know, at least one and a half years, we're going to see increases in terms of these Systems capabilities. So AlphaCode1, you could see was at 50%. This one right here was at 85%. And then we can see in the future, I think it's going to be earring near towards 90%. And then we might even have agents from OpenAI that even dwarf these capabilities. So they. This is something that's rather important to pay attention to because once these kinds of systems are deployed, how are they going to change the software engineering landscape? How are they going to change the coding landscape? And I think AlphaCode 3, AlphaCode 4, I mean, with these kind of trends, I think it's pretty clear where certain things are going. If you are in the industry now, something that did get a lot of flack was Devon. But I've made the point and a lot of people did agree with me when I did make this point. Now, for those of you are software engineers or are budding coders, you might argue that, yeah, this was faked or whatever. But that's not my argument here. My argument here is that, look, let's take a look at what Devin actually is. So Devin is essentially your first AI software engineer. And I guess you could say we could put that in quotation marks. But this isn't the crazy thing, okay? The crazy thing about this is of course, the benchmarks. So if I highlight the benchmarks, I'm not sure exactly where they are. They should be around here. You can see that this shows us a Stark improvement from GPT4 to Devon. Now, essentially, Devin is evaluated on a random 25% subset and the results are unassisted where all other model results are assisted. But the craziest thing about this is that Devin is literally just a GPT4 wrapper. I'm not entirely sure what Devin have done. They haven't released a technical paper, but the gist of things is that they've managed to introduce some kind of framework that when they wrapped GPT4, I'm guessing they can get up to 13.86% and that allows them to receive a huge boost in terms of the software engineering capabilities. Why is this such a huge red flag for anyone trying to enter that industry at the time? Well, once again, like I said before, we need to take a look at where trends are headed. We can see that things have jumped from 1.4% all the way up to 13.86%. Now, if you don't get what I'm getting at here, basically what I'm saying is that this is just a GPT4 wrapper. I'm not betting on people working at Devon to make this any better. What I'm betting on is GPT5. Sam Altman has previously stated that you don't want to bet against their company because they're going to steamroll your AI company. You should instead try to build things with the fact that GPT5 is going to get better. And essentially this is exactly what devin is. Once GPT5 gets a lot better and solves many of the problems that currently exist, I'm arguing that Devon is probably going to receive another massive boost in performance, and thus performance in the software engineering bench, which then again means that we're going to see an increased impact in terms of that industry. Now, I'm not stating that it's going to jump to 99%, incredibly, but we truly never know how smart this model is going to be. And the way that I've heard OpenAI employees and members talk about that GPT5 system and the way how they talk about GPT4 as if it's this, you know, really, really unintelligent thing that just doesn't understand what's going on, it leads me to believe that successor systems like GPT5 and then of course, GPT6 will be remarkable improvements on Devon and whatever other systems are going to be there in terms of the coding capabilities, which means that in the future, coding is likely to be nearly all automated at that point, which means that if we look at the trends right now, it might not be the best time to actually start that career. Now, there's also something else. Remember how I said in the future, things that we're looking at are going to get increasingly, increasingly better? And the problem is, is that a lot of people are now starting to realize how much money there is floating around the generative AI space. If you don't know, Devin, once they released this demo, I. I think they had a $25 million funding round and they've been getting more and more funding as many people have piled onto the train. And recently there was this company called Magic that recently got $100 million investment. Basically, the gist of this, I'm going to explain it to you really quickly. The guy, Nat Friedman, essentially saw a demo from the people who were starting this company, and the demo shocked him that much that he said, you know What? I'm putting $100 million into this and the rest is history. So they haven't announced their demo yet. I'm guessing that they're currently probably working on beating the benchmarks that were shown in Devon. But what they showed us, the advanced level of reasoning, the fact that someone just initially invested based on a short demo, shows us that the industry is truly about to be shaken up. Which means that the future for software engineers and for those in coding is definitely quite uncertain. And it doesn't make sense to go into an industry that's quite uncertain. So that is something that I would argue that just doesn't currently make sense.

The amount of change we're going to see over the next five years, 10 years, will dwarf everything that's happened over the last 30. We went from automating pen and paper, right, with spreadsheets, to connecting PCs into local area networks, to connecting them. Connecting people locally. Yeah, locally, right. To connecting networks to get the Internet, to getting the network effect of communicating people globally. But now we're introducing machine intel, you know, artificial intelligence, machine learning, deep learning, neural networks, etc. So. And what that enables is the automation of automation, right? And so the people who were writing software, particularly at the lower end, unless you. You're doing these advanced things, they're gone, right? The people that software's writing itself, it's doing itself, right? It's just math. Programmer back in the day. So when I was writing code, it was the algorithms were if this, then that, right? Xor, whatever it may be, right. But you had to guess, right, you had to use your best instance and then it got smarter and smarter and smarter and libraries to do bigger and better things. All that is being automated. And so now you have to know how to use that stuff. So either software works for you or you work for software. And once the software takes over, you're gone. So unless you understand that, you don't understand the nature of work is changing, the nature of employment is going to change. And from a business, from a stock perspective.

So that was Mark Cuban actually talking about software engineering. And we can clearly see that he makes some valid points. You know, it's basically an inflection point where things are changing on a truly, truly pivotal moment. Because this is the automation of automation. Unlike the invention of new tools, this is a tool that doesn't really need anyone to oversee it or anyone else to sort of manage it. This is a tool that can pretty much run autonomously and that is the entire goal of AI. It's automation, artificial intelligence. And with the fact that this could be increasingly better in the future and you could have systems that could run code by themselves, they could be self repairing. And I know that does sound like sci fi, but let's not say that, because we've seen some crazy things this year. It increasingly begs the question, would you now start off in this industry with those looming careers prospects? In addition, there are some things that state there is actually quite a bleak outlook in terms of the coding area. If you look at the tech companies, the ones that hire and pay the biggest salaries, we know that there are a bunch of layoffs Continuing into 2024, major layoffs at Tesla, Amazon and Microsoft from small fintech startups and apps. We know that there isn't as much money floating around in certain areas anymore. And the days are gone where you could just do a coding boot camp and get a 200 to 300k a year salary and just relax whilst making high six figures. Things are really, really ramping up. And if you don't believe me, this is a recent tweet where someone's actually talking about ubi. He says, I used to be against universal basic income, but it's become inevitably, increasingly clear that the population is wildly over educated and underemployed. He said 50% of the workforce is about to lose to AI and robotics. And we can see here that 12,000 people have applied for a $20 an hour software engineering internship. That is something that is pretty, pretty insane. So I guess you could say that right now things are rather saturated and everyone is over educated. Because if you do remember, there was that meme ages ago where people were saying learn to code and now it seems like that has flipped against them. Now I'm going to argue the other side because whilst yes, everything that I just did is probably going to give you the comprehensive picture that, you know, why on earth would I learn to code if you know, AlphaCode 3 is going to literally destroy my job. If Devin just gets a boost with GPT5, it's automatic. Estimated there are these companies pumping hundreds of millions of dollars. You know, CEOs of major companies and billionaires are saying, look, don't even bother learning to code because that's going to be gone in the future. And with an increasingly bleak outlook and rapid oversaturation and competition, why on earth would you. Well, let's actually take a look at exactly what's going on. So at this part of the video, I'm going to present to you why you probably should still learn to code. I know you might be thinking, why on earth would you do that? Guys, it's always important to have both perspectives because if you don't, then you're not thinking about the problem and you leave yourself open to blind spots. So one of the things that is really, really true, and I've spoken about this in another video, is that whilst, yes, the successor systems are going to be increasingly better in terms of their reasoning accuracy and reliability, the problem is, and with AlphaCode 2, whilst, yes, it was better than 85% of people we can see from the research paper, despite AlphaCode 2's impressive results, a lot more remains to be done before we see systems that can reliably reach the performance of the best human coders. Our system requires a lot of trial and error and remains too costly to operate at scale. Further, it relies heavily on being able to filter out obviously bad code samples. So if you don't know how AlphaCode basically works, it works like this. So basically they generate a bunch of different codes, they filter out the bad ones, analyze the good ones, then they test them and then whichever one works, that's the one they use. And then the thing is, they don't do this at a small scale, they do this at a giant scale. So using all of that, compute, they're basically brute forcing every single opportunity and then narrowing it down to the few that actually work. Right now the problem is that that is too costly to scale. So this is something that is far away from production because it means that either A they're using the wrong architecture because they're just trying to brute force it, or either B we just don't have enough energy for this. And even if that is the right architecture, it means that it's going to be a while before this does come out. So this is something that isn't currently reliable. So with AlphaCode 2 there are still some limitations, which means it's going to be some time before we do get that system. Another reason is the fact that whilst yes, Devin was pretty crazy, this kind of has been debunked. If you haven't seen this video, I'm pretty sure you have. It basically dives into how the first software engineer There were some little bit of exaggerations and I guess we can contribute those exaggerations to the fact that software engineering is a pretty intense industry and with all these companies sprouting up, a lot of people want a piece of the AI pie and in doing so they're going to exaggerate their demos as much as possible to create hype and buzz around their industry and of course their companies. So there may have been some slight exaggerations with their jobs, such as the Devon demo completing an upwork task that was quite misleading or outright fabricated. And the code that it made was described as hard to maintain and modify by some engineers. And the benchmark is only 13.86% which can't really replace any human level developers. And the fact that this demo from Devin was actually sped up quite a bit, it apparently completed the tasks in 24 hours that would take the average software engineer around one or two hours to do. So that is something that is pretty crazy. And of course I'm guessing the inference cost of this isn't justifiable just yet. Now in addition, there are also some transferable skills. In 2024, learning to code could still be highly valuable as it can be applied to the rapidly expanding field of machine learning and artificial intelligence. And here are some reasons why. So AI and machine learning actually rely on coding foundations. So whilst AI and machine learning are becoming more accessible, they still rely heavily on coding skills. Many AI and machine learning tools and frameworks such as TensorFlow, PyTorch and SciKit Learn are built on programming languages like Python. Understanding coding allows you to effectively utilize these tools, customize them for your needs and troubleshoot issues. Coding enables deeper understanding of AI and machine learning to code. Coding enables a deeper understanding of AI and machine learning And Learning to Code provides deeper understanding of how AI and machine learning algorithms work under the hood. And by implementing these algorithms from scratch, you gain insights into their inner working and their assumptions and their limitations. And this is crucial knowledge for designing effective AI and machine learning solutions. In addition, we do have the fact that there could be increased and continued demand for skilled software engineers. Despite the major concerns about AI and automation, the demand for skilled software engineers is projected to remain strong in the coming years. The U.S. bureau of Labor Statistics actually predicts a 22 to 26% increase in employment over the next decade. However, I got to be honest with the caveat. I'm not sure if they're accounting for artificial general intelligence or some of the technology that has been recently developed. We can also see here the demand for software engineers had doubled since 2020. The negative effects of the COVID 19 pandemic was rather short lived as the software market value is expected to continue growing at an annual rate of 6.5%. And we can see here that these companies in terms of their revenue is increasing in terms of billions and billions of dollars. And in addition we have the continued demand of SaaS companies increasing. We can see many software engineers start their careers in SaaS companies. The industry is booming in part because the recent pandemic made many companies go remote and increase their cloud spending as a result. And expected to reach 208 billion by 2023. And the point is, is that this is an industry that is continuing to expand. Whilst yes, you could argue that they might need newer, less developers, it's still an expanding industry. Now you could also state that AI is expected to augment and enhance the work of programmers rather than replace them entirely. Because AI powered tools can assist programmers automatically generating code, snippets, functions or even entire modules based on high level specifications or patterns. And this can pretty much save you time and effort in writing repetitive or boilerplate code, allowing programmers to focus on more complex and creative tasks. Which means that the industry is just changing and not going away. Now overall, if we look at things in a 15 year time horizon, I would argue that maybe it might not be the best time to code. But the problem with this is that making predictions right now is a fool's game. Because it's so hard to think about what's going to happen. Because the problem is, is that many have described the arrival of AGI as a singularity where nobody knows what's going to happen. So right now it's pretty, pretty hard. I would always evaluate my job prospects on a continued basis and I would always say whatever industry you're in, just make sure I can give this advice to you. Just make sure you're paying attention. If there's a new tool, if there's something new that could impact your industry, make sure you're one of the first people to find out because usually if you're the last to find out, it's usually too late to make any meaningful changes. Imagine like two years from now you find out about Devon. Like by that time I'm pretty sure if you're going to make any kind of change, it's going to be too late. So the point is that whilst, yes, the pathway does seem a little bit bleak, I would say pay attention to the industry and if you see that there are some career opportunities that are going to be vastly improving and, and areas where you could potentially apply your skills or have transferable skills, definitely take a look in those areas because it's important to pay attention.